Attention Based Hybrid Deep Learning models for Multi class Amharic News Categorization with Explainable AI

Attention Based Hybrid Deep Learning models for Multi class Amharic News Categorization with Explainable AI | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Attention Based Hybrid Deep Learning models for Multi class Amharic News Categorization with Explainable AI Amlakie Aschale Alemu, Malefia Demilie Melese, Daniel Arega Mengesha, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6759877/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 11 You are reading this latest preprint version Abstract Efficient and adaptable text classification systems are required due to the increasing expansion of Amharic electronic documents in order to manage and extract important insights from large amounts of data. Traditional natural language processing (NLP) methods are severely hampered by the Amharic language's intricate morphology and sparse annotated corpora. By combining the advantages of Bidirectional GRU and Bidirectional LSTM networks for capturing sequential dependencies and Convolutional Neural Networks (CNN) for local feature extraction, this study proposes a novel attention-based hybrid deep learning model for multi-class Amharic news classification. The model's discriminative power across several categories is improved by incorporating a self-attention mechanism that dynamically emphasizes contextually significant terms. We employ Explainable AI (XAI) methods, like Local Interpretable Model-Agnostic Explanations (LIME), which offer human-interpretable explanations for predictions, to minimize the hidden nature of deep learning decisions and enhance trust in AI systems. A carefully selected multi-class Amharic news dataset that encompasses a variety of topics, including agriculture, culture and tourism, education, the economy, the environment, foreign affairs, health, politics, science and technology, and sports, is used to evaluate the model. The experimental results show that CNN + BiLSTM with Self-Attention scores 96.7, 96.8, 96.8 and 97 for precision, recall, F1-score, and accuracy respectively. This study introduces to the larger objective of transparent and reliable AI in low-resource language contexts and pushes beyond the limits of Amharic automatic language processing. Explainable AI Self_Attention CNN + BiGRU CNN + BiLSTM LIME Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. Introduction Low-resource languages like Amharic, which is one of the most frequently spoken languages in Africa and Ethiopia's official working language, have also been affected by the exponential growth in digital content worldwide [1], [2], [3],[4]. Automatic classification of this data is essential for sentiment analysis, information retrieval, focused on news delivery, and effective information framework [5]. However, because of its rich morphology, complicated syntactic structure, and absence of annotated corpora and computational resources, Amharic presents particular difficulties for natural language processing (NLP) [6], [7], [8], [9]. Support Vector Machines (SVM) and Naïve Bayes are two examples of traditional machine learning techniques for text categorization that mostly rely on human feature engineering and shallow representations like bag-of-words or TF-IDF, which frequently fall short in capturing the semantic and contextual richness of texts [10]. Whereas end-to-end learning of hierarchical features made possible by deep learning models has transformed natural language processing (NLP) [6], [9], acceptance of them in Amharic and other low-resource languages is still quite restricted [9], [10]. When attempting to do multi-class classification [11], [12], [13], [14], [15], [16] of Amharic articles, we present a hybrid deep learning architecture that employs Convolutional Neural Networks (CNNs) with Bidirectional GRU (CNN + BiGRU) and Bidirectional LSTM (CNN + BiLSTM) networks, enriched with a self-attention mechanism. While BiLSTM layers model long-range dependencies and contextual flow in both forward and backward directions, CNN layers are used to extract local n-gram features that capture syntactic patterns [17], [18], [19]. The model can concentrate on contextually relevant tokens that are required for classification because the attention mechanism dynamically weights the importance of various words in a sequence [15], [20], [21], [22]. Furthermore, one of the most major drawbacks of deep learning models is their black-box dimension, which makes them difficult to interpret and undermines user confidence, particularly in low-resource environments where linguistic proficiency is crucial [23], [24], [25]. We use Explainable AI (XAI) techniques in our framework to tackle this. While techniques such as Local Interpretable Model-Agnostic Explanations (LIME) are used to provide after the fact model explanations, attention visualization specifically emphasizes the role of individual tokens in the final prediction [26], [27], [28]. A newly developed or designed multi-class Amharic news dataset with annotations across a variety of topical groups is used to test the model that we propose. According to experimental evaluations, our hybrid architecture performs better in terms of classification accuracy, macro F1-score, and robustness than both standalone deep learning models and conventional baselines. Likewise the model's ability to offer meaningful and intelligible insights into the categorization reasoning is verified by attention corresponds and LIME-based explanations [29], [30]. By (i) presenting an efficient deep learning pipeline designed for the Amharic language, (ii) resolving the interpretability issue with integrated XAI techniques, and (iii) offering a useful standard for further research on Amharic news classification and multilingual explainable NLP systems, this study makes a major contribution to the field of NLP for low-resource languages. The primary contributions of the study are as follows. We propose a hybrid model that combines Convolutional Neural Networks (CNN) with Bidirectional GRU (CNN + BiGRU) and Bidirectional LSTM (CNN + BiLSTM) networks, effectively capturing both local semantic patterns and long-range dependencies in Amharic news texts. This architecture is further enhanced by a self-attention mechanism, allowing the model to selectively focus on the most informative words in a sequence. The model is specifically designed and optimized for the Amharic language, addressing its unique linguistic characteristics such as rich morphology, syntactic variations, and limited annotated datasets. This research bridges a critical gap in deep learning applications for low-resource languages. We integrate Explainable AI (XAI) techniques into the classification framework. Local Interpretable Model-Agnostic Explanations (LIME) offers post-hoc explanations. Together, these methods enhance transparency and user trust in the system’s decisions. A multi-class labeled dataset of Amharic news articles covering categories such as politics, economy, health, sports, and technology is compiled and annotated. This dataset contributes to the Amharic NLP community and serves as a benchmark for future research. This work contributes to the broader effort of democratizing NLP technologies by advancing research in low-resource African languages. It demonstrates that deep learning models, when properly adapted and explained, can perform competitively even in resource-scarce contexts. 2. Related Works The paper in [5] provides an in-depth examination of XML approaches used in text classification models, including post-hoc interpretability frameworks such as LIME and SHAP, attention mechanisms, and feature significance strategies. Additionally, by balancing the trade-offs between interpretability and performance, they evaluate these methods' ability to improve model transparency. The study also explores how XML may develop in text classification in the future, including how user-focused explanations and legal requirements for AI transparency could get incorporated. According to the findings, explainability helps to detect model biases, improve overall performance, and increase the trustworthiness of the model. The work [28] provides an integrated strategy for fake news detection, applying three publicly available datasets: WELFake, FakeNewsNet, and FakeNews Prediction. They merged FastText word embedding with various Machine Learning and Deep Learning methods, further refining these algorithms with regularization and hyperparameter optimization to prevent overfitting and enhance model generalization. In addition, they utilized state-of-the-art transformer-based models like as BERT, XLNet, and RoBERTa, improving them through hyperparameter adjustments. In the concluding phase, explainable AI modeling was applied using Local Interpretable Model-Agnostic Explanations, and Latent Dirichlet Allocation to get deeper insights into the model’s decision-making process. The work [31] explores the integration of Explainable Artificial Intelligence (XAI) methodologies. This study's main goal is to assess how various XAI techniques can be used to interpret and explain the choices made by NLP-based misinformation classifiers. To investigate model behavior and feature importance, XAI techniques such as SHAP (SHapley Additive Explanations), LIME (Local Interpretable Model-agnostic Explanations), and attention visualization were used. The results show that XAI techniques greatly increase transparency by highlighting key words and phrases that influence model decisions. The study [32] explores how sentiment analysis, utilizing data from social media platforms like YouTube, can improve thier comprehension of how the public responds to HMPV. It emphasizes how crucial it is to monitor public state of mind, which ranges from fear to trust, in order to inform regulations, address false information, guide health communications, and promote adherence to preventive measures during outbreaks. In sentiment classification tasks, the study uses sophisticated transformer models, especially XLNet, and achieves an accuracy of 93.50%. Furthermore, they apply SHAP to integrate explainable AI (XAI) to make it transparent how the model determines the main determinants of public perception. In study [33] provides a new deep learning-based method for detecting misleading content on social media using the sine cosine algorithm (SCADL-DCDSM). To classify emotions, the SCADL-DCDSM technique uses a hyper-parameter tuning strategy combined with the ensemble learning process. Furthermore, three models long short-term memory (LSTM), extreme learning machine (ELM), and attention-based recurrent neural network (ARNN) are used in the SCADL-DCDSM technique for sentiment classification. Lastly, SCA can be used to improve the DL models' hyper-parameter selection. In order to improve accurate deceptive content recognition, the SCADL-DCDSM system incorporates the explainable artificial intelligence (XAI) system LIME, which has been used for an extensive explainability and grasp of the black-box process. The results of the model showed that, when considering several metrics, the SCADL-DCDSM methodology produces the best result. According to study [34] human beings are able to evaluate machine learning models for accuracy, fairness, and dependability due to the developing discipline of explainable artificial intelligence (XAI). Using a well-known Explainable AI method called local interpretable model agnostic explanations (LIME), they build on our earlier research and conduct a thorough investigation of the model developed for text categorization and sentiment analysis. The findings confirm that explainable AI tools are necessary for evaluating machine learning models because accuracy and other associated metrics do not guarantee the model's correctness, fairness, and dependability. Furthermore, they use LIME to compare the interpretability and explainability of different machine learning techniques. A feature fusion framework based on the bidirectional encoder representations from transformers (BERT) is proposed in paper [35]. This hybrid approach trains the word vector representation using BERT. Static characteristics are extracted by convolutional neural networks (CNNs). Furthermore, to gather contextual data, a bi-gated recurrent neural network (BiGRU) is used. In addition, an approach for determining the weight for important words is introduced. The proposed approach performs noticeably better than the other advanced baseline techniques, according to the experimental data. In the paper [36] they presented a hybrid LIME-BiLSTM model in which LIME ensures explainability and transparency while BiLSTM provides classification accuracy. The proposed framework becomes understandable in this integrated model since LIME explains the prediction and performs similarly to the original model. Kendall's tau correlation coefficient is used to assess this model's explainability performance. likewise they use a number of machine learning models and compare how well they work. Because the model adopts an integrated approach, they examined and contrasted the computational overhead of our suggested model with the other approaches. The study in [37] offers a multi-class approach to detecting fake news that combines Gradient Boosting methods with Transformer-based deep learning models. Furthermore, interpretability and insights into model predictions are provided by Explainable AI (XAI) techniques as SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations). The experimental findings, which were obtained using benchmark datasets like LIAR and FakeNewsNet, show that the proposed mixed approach performs better than standalone models in terms of interpretability, accuracy, and F1-score. A major discrepancy in the evaluation of eXplainable Artificial Intelligence (XAI) techniques for text categorization is addressed in the research [38]. While current frameworks concentrate on evaluating XAI in domains like visual analytics and recommender systems, an extensive evaluation is missing. Recent afterwards XAI techniques are surveyed and categorized in our study based on their output format and explanatory scope. The main conclusions show that feature-based explanations are more accurate than rule-based ones. These revelations improve knowledge of XAI in text classification and provide insightful advice for creating successful XAI systems. Table 1 shows how our study's model comparison utilizing accuracy is depicted. Table 1: Comparison of the proposed model with existing state of the art models Article References Tasks Model Model Accuracy Applying Attention Mechanism Applying Model Interpretability [4] Categorization CNN 93.79% No No [39] Classification CNN+BiLSTM 91.41 Yes No [40] Classification BiLSTM 88.00 Yes No [41] Classification CNN+BiLSTM 96.08% No No [42] Classification BiLSTM 96.23 No No [43] Classification LSTM 92.34 Yes No [44] Classification CNN+BiLSTM 84 No Yes Our study Categorization CNN CNN+BiGRU CNN+BiLSTM 90 92.7 97 Yes Yes 3. Methods and Materials The following figure 1 illustrates a comprehensive end-to-end pipeline for Amharic text news classification, starting from data acquisition to model interpretation. The proposed approach for multi-class Amharic news classification follows a structured pipeline that integrates data preparation, hybrid deep learning model construction, and explainable AI mechanisms. 3.1. Amharic Text News Data Collection: Throughout this stage, the raw Amharic articles are collected from various digital sources and online news sites. To ensure the diversity and the representativeness of the categorization task, the data covers a wide range of areas, including agriculture, culture and tourism, education, economy, environment, foreign affairs, health, politics, science and technology, and sports. Since there doesn't appear to a publicly available Amharic document classification corpus, we collected 1200 Amharic news documents from ten major news categories, each news category containing 1500 sentences, in order to train and test the deep learning model. 3.2. Data Cleaning and Preprocessing: Noise, inconsistencies, and unnecessary characters like punctuation, stop-words, numbers, and symbols are frequently found in collected text data. This process involves the following steps: (1) Unicode standardization and text normalization; (2) removal of non-Amharic tokens, special characters, and redundant white spaces; and (3) normalizing of common morphological variants to ensure consistency. After the fact that the data is cleaned and preprocessed, which includes normalizing the text for consistency and removing noise like punctuation, numbers and non-Amharic characters. To get the text prepared for analysis, this step may include lemmatization of stemming, and stop word removal. 3.3. Data Labeling and Tokenization: Data labeling and tokenization come next after preprocessing. In order to get ready the text for machine learning input, each article is given a label that corresponds to its category (politics, sports, business, etc.). The text is then tokenized and divided into smaller chunks, such as words or sub-words. Every news item in this stage is given the appropriate class label. Splitting texts into linguistically meaningful units (tokens), usually words or sub-words, is known as tokenization. To transform textual tokens into dense vector representations appropriate for deep learning models, word embedding tools like FastText are optionally used here. 3.4. Dataset Splitting: To ensure equal representation across categories, the dataset is divided into training and testing sets. This stage makes it easier to learn the model and evaluate it accurately. Training and testing are done in a standard ratio, like 80:20. The labeled dataset is subsequently split into subsets for testing and training, usually known as dataset splitting. We employed a total of 2200 datasets, of which 1760 documents were used for training and the remaining 440 documents for testing. This makes it possible to test the model's generalization performance on one set of data and train it on another. 3.5. Model Building (Hybrid Deep Learning with Self-Attention): Deep learning architectures are used in the pipeline's basic process of building models. CNN (Convolutional Neural Network), CNN blended with BiGRU (Bidirectional Gated Recurrent Unit), and CNN integrated with BiLSTM (Bidirectional Long Short-Term Memory) are among the models that have been highlighted. In order to better comprehend contextual meaning in Amharic text, these models are further improved with self-attention mechanisms that enable them to concentrate on the most pertinent portions of the text independent of word position. An entirely novel hybrid architecture develops that includes: CNN (Convolutional Neural Networks) to capture local spatial features and patterns in the text. Bidirectional GRU (CNN + BiGRU) and Bidirectional LSTM (CNN + BiLSTM) to model long-range and bidirectional dependencies in sequences. A self-attention mechanism is integrated to weigh the contextual importance of tokens, allowing the model to focus on semantically relevant parts of the text. This hybrid structure combines the strengths of both convolutional and recurrent layers, enhanced by attention to improve classification performance and interpretability. A hybrid deep learning model which includes Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM) networks, and a Self-Attention mechanism for text classification is depicted in the architecture figure 2. There are multiple logical sections to this model's structure: A preprocessed text dataset that has undergone common cleaning techniques including tokenization, stop-word removal, and normalization is forwarded to the input phase. The word embedding section accepts this processed data and uses pre-trained FastText word embedding to turn each word into a dense vector. By using subword information to determine both the morphology and meaning of words, these embeddings enhance the representation. Beyond that, the embedded sequences get passed to the feature extraction section, which uses a number of 1D convolutional filters to identify local patterns in the text, like n-grams. By emphasizing grammatical structures, the convolutional layers serve as effective feature extractors. Following that, max-pooling layers increase efficiency and generalization by reducing the dimensionality of the feature maps while keeping the most noticeable features. Bidirectional LSTM (BiLSTM) networks are used in the sequence learning section to identify long-range dependencies in the text. The BiLSTM outputs are subsequently submitted to a self-attention layer via the attention mechanism section. By using this technique, the model can focus on informative terms and eliminate out irrelevant ones by allocating weights to various tokens according to their significance to the classification task. By attending to critical words across the entire sequence, the self-attention module enhances the interpretability and effectiveness of the model. The attention layer's enhanced characteristics are subsequently processed via a dense layer with ReLU activation and a softmax output layer in the classification stage. The softmax layer provides class probabilities, which show the model's prediction for every input sequence, while the ReLU layer offers non-linearity, allowing the network to learn intricate patterns. This CNN-BiLSTM-Self Attention model is especially well-suited for sophisticated natural language processing tasks including sentiment analysis, document categorization, and intent identification because it effectively determines local patterns, long-term dependencies, and global contextual relevance. During training, the model is represented as the trained model, which can now predict new, unseen text. 3.6. Model Evaluation: The next step involves model evaluation, where performance metrics such as accuracy, precision, recall, and F1-score are used to assess how well the model performs. The model is trained on the labeled training dataset using appropriate loss functions and optimization algorithms. Evaluation is conducted on the test set using standard classification metrics. In this work, we evaluate the proposed models' performance using evaluation measures like accuracy, precision, recall, and F1 score which are formulated as the following equations 1-4. Where, TP, TN, FP and FN represent true positive, true negative, false positive and false negative respectively. These metrics provide insight into the model's effectiveness across all news categories. 3.7 Model Interpretability with Explainable AI (XAI ): Subsequently, the pipeline blends model interpretability through the use of tools such as LIME (Local Interpretable Model-agnostic Explanations), which improve transparency and reliability by offering insights into the model's decision-making process. LIME (Local Interpretable Model-Agnostic Explanations) is a subsequent interpretability tool that produces human-readable explanations for individual predictions in order to overcome the black-box feature of deep models and increase user trust. By ensuring transparency, this element assists in the understanding and validation of the system's behavior by users and domain experts. All things taken into account, this process integrates preprocessing, advanced neural network models, and interpretability techniques to produce a reliable and understandable system, representing a comprehensive and contemporary approach to Amharic text classification. The proposed technique for classifying multi-class Amharic news uses a structured workflow which integrates the building of hybrid deep learning models with data preprocessing. 4. Results and Discussions To compare convolutional recurrent neural networks with distinct characteristics and assess the study's performance in regard to the challenges we looked at, this section offers the built models and the experiments with the results collected. We must analyze the dataset we used to build our model in order to accomplish this. The outcomes produced by the implementation are investigated, and the metrics are used to assess the results. To find the one that would work best, we experimented with several combinations of the model's hyper-parameters. Thus, as indicated in table 2, we select the following hyper-parameters. Table 2: Hyper-parameter configuration details for the proposed model Hyper-parameters Models Epoch Batch size Learning rate Optimizer Activation Dropout CNN 12 32 1e-5 ReLu SoftMax 0.6 CNN+BiGRU 12 32 1e-5 ReLu SoftMax 0.6 CNN+BiLSTM 12 32 1e-5 ReLu SoftMax 0.6 Generally, when looking at the whole hyper-parameters and models of the experiments, table 2 gives a short and precise analysis of the experimental result of this study for the selection of hyperparameters. Table 3: Evaluation metrics of deep learning models Attention mechanism Evaluation Metrics Models Precision Recall F1_Score Accuracy CNN 89.8 89.9 89.7 90 CNN+BiGRU 92.4 92.3 92.4 92.7 CNN+BiLSTM 96.7 96.8 96.8 97 Three deep learning models CNN, CNN+BiGRU, and CNN+BiLSTM are compared in the evaluation metrics which is plotted in table 3 based on four important classification metrics: accuracy, precision, recall, and F1 score. The baseline model, the Convolutional Neural Network (CNN), achieves an overall accuracy of 90, an F1 score of 89.7, a precision of 89.8, and a recall of 89.9. We notice considerable improvements in every metric when a Bidirectional Gated Recurrent Unit (BiGRU) is integrated to the CNN. The accuracy, precision, recall, and F1 score of the CNN+BiGRU model are 92.2, 92.4, and 92.3, respectively. The integration of BiGRU, which enables the model to account for both past and future context in the input sequences a vital feature in tasks like natural language processing is responsible for this enhancement. The CNN+BiLSTM model provides the most notable performance. Out of the three models, this combination produces the highest accuracy (97), precision (96.7), recall (96.8), and F1 score (96.8). The enhanced outcomes indicate that BiLSTM is even more successful than BiGRU at figuring out long-term dependencies in the data. When everything is taken into account, the performance trends render it abundantly evident that blending CNN with recurrent architectures improves the model's grasp of sequential or contextual patterns in data. CNN+BiLSTM consistently perform better than CNN+BiGRU among the two hybrid models, despite at a possibly better computational cost. CNN+BiLSTM were the proposed alternative for applications requiring the highest level of accuracy and robustness. Even though it is less precise, the standalone CNN model might nevertheless work well in contexts with limited resources or a lightweight. Three deep learning architectures are compared in Figure 3 based on four common classification metrics: accuracy, precision, recall, and F1 score. CNN with Self-Attention, CNN with BiGRU and Self-Attention, and CNN with BiLSTM and Self-Attention are among the models. With values nearly 97, the CNN+BiLSTM+Self-Attention model outperforms than the other two models in every criterion, indicating higher classification capabilities. According to this, introducing bidirectional LSTM layers greatly improves the model's capacity to identify textual sequential dependencies, which are then further refined by the attention mechanism. Considering metric values ranging from 92 to 92.3, the CNN+BiGRU+Self-Attention model shows a moderate improvement over the CNN+Self-Attention baseline, indicating that GRU-based sequence modeling provides a useful improvement over convolutional layers alone. The CNN+Self-Attention model, on the other hand, performs comparatively poorly in grasping contextual complexities, even if it is still robust. Its scores are just below 90. The efficiency of integrating sequential neural layers with attention mechanisms to improve text classification tasks in complicated language datasets has been shown by a gradual rise in all four measures across model configurations. Table 4: Classification Report for CNN+Self Attention When recurrent and attention mechanisms are integrated, the models' classification performance clearly shows an upward trend as indicated in table 4. With a 90% overall accuracy and weighted and macro averages of precision, recall, and F1-score of 0.90, the CNN+Self-Attention model offers a strong foundation. The model's performance is marginally worse in the areas of Economy (F1-score = 0.79) and Politics (F1-score = 0.72), but it exceeds in Sport (F1-score = 0.99) and Science and Technology (F1-score = 0.95). These differences imply that CNN by itself does a good job of identifying patterns in some domains, especially when the terminology or textual structure is consistent, but it may have difficulties with more complex or context-heavy categories. Table 5: Classification Report for CNN+BiGRU+Self Attention In Table 5 the model's contextual understanding is greatly enhanced by incorporating a BiGRU layer in along with CNN and Self-Attention, increasing overall accuracy to 92%. The precision, recall, and F1-score weighted and macro averages all rise to 0.92. In comparison to the former model, this model shows noticeable enhancements in Politics (F1 = 0.86) and Economy (F1 = 0.84), while it achieves almost perfect performance in classes like Education (F1 = 0.98) and Sport (F1 = 1.00). The BiGRU improves the model's comprehension of sequential interactions in both directions, which is particularly useful in fields like foreign affairs and health that need for a more thorough contextual understanding. Table 6: Classification Report for CNN + BiLSTM + Self-Attention In Table 6 the CNN+BiLSTM+Self-Attention model performs the best, accomplished remarkable accuracy of 97% and weighted and macro averages of 0.97 for all measures. In nearly every field, this model performs admirably; in fact, certain classes (including education, science and technology, and sport) have perfect F1-scores of 1.00. With F1-scores of 0.96 and 0.91, respectively, even formerly poorer classes like Economics and Politics demonstrate significant progress. When combined with self-attention's capacity to concentrate on important parts of the sequence, the BiLSTM's ability in collecting long-range interdependence appears to be particularly useful for activities requiring both local and global contextual awareness. In conclusion, adding recurrent layers and attention mechanisms to CNNs leads to marked improvements in classification performance. The progression from CNN+Self-Attention to CNN+BiGRU+Self-Attention, and finally to CNN+BiLSTM+Self-Attention, reflects increasing model complexity and interpretive capacity. To evaluate the effectiveness of the proposed attention-based hybrid deep learning models for multi-class Amharic news classification, we analyzed the confusion matrices generated for three variants: CNN with BiLSTM and self-attention, CNN with BiGRU and self-attention, and CNN with self-attention only as plotted in figure 4. These matrices provide insights into the models' ability to correctly classify instances across various news categories, as well as their tendency to confuse certain classes. The CNN + BiLSTM + Self-Attention model demonstrates the most robust performance among the three as indicated in Figure 4 subfigure (a). It achieves high accuracy across most categories, with particularly strong classification in classes such as Economy, Education, Science and Technology, and Sport, showing correct predictions of 54, 38, 52, and 41 respectively. The model also exhibits minimal misclassifications, which highlights its strength in capturing long-range dependencies and contextual information within Amharic text. However, minor confusion is observed between semantically related categories such as Politics and Foreign Affairs, which is expected given the overlapping nature of their content in real-world news reporting. In Figure 4 subfigure (b), the CNN + BiGRU + Self-Attention model also performs well, showing strong results in categories like Agriculture, Environmental, Health, and Sport. While slightly less accurate than the BiLSTM-based model, it maintains a competitive level of performance while being more computationally efficient. Nevertheless, a slightly higher degree of misclassification is noted in categories such as Politics, Culture and Tourism, and Education. This can be attributed to the lighter architecture of GRUs, which, while faster, may be less expressive in capturing complex sequential patterns compared to LSTMs. In Figure 4 subfigure (c), illustrated CNN + Self-Attention, which excludes the recurrent component, achieves reasonable accuracy but exhibits noticeable limitations in modeling complex class relationships. It performs well in relatively distinct categories such as Sport, Health, and Science and Technology, where the contextual boundary between classes is clearer. However, its performance drops in categories like Politics and Economy, where subtle contextual differences are crucial for accurate classification. This indicates that while self-attention enhances the model's ability to focus on relevant tokens, the absence of sequential modeling (as provided by RNNs) reduces its effectiveness in capturing deep contextual meaning. In general, the analysis confirms that the integration of CNN, bidirectional recurrent layers (BiLSTM or BiGRU), and self-attention yields significant improvements in classification performance. Among the models evaluated, the CNN + BiLSTM + Self-Attention configuration consistently provides the most accurate and reliable results, particularly in classes that require understanding of long-range dependencies. Figure 5 presents a comparative ROC curve analysis of three different classification models evaluated on a multi-class news categorization task. Each plot illustrates the class-wise ROC curves and their corresponding AUC values across ten thematic categories: Agriculture, Culture and Tourism, Economy, Education, Environmental, Foreign Affair, Health, Politics, Science and Technology, and Sport. (i) Left ROC Plot ( Figure 5 sub Figure (a) ): With Culture and Tourism having the highest class-wise AUC (0.61), this plot shows comparatively balanced AUC values for the majority of categories, indicating superior separability in this region. On the other hand, Foreign Affair (AUC = 0.45) and Education (AUC = 0.36) show lower predictive ability, indicating that it may be challenging to correctly differentiate these groups from others. Limited discriminating power and possible model under-fitting are highlighted by the AUC values, which primarily range around 0.50. (ii) Middle ROC Plot ( Figure 5 sub Figure (b) ): AUC values in this configuration are generally lower than in the preceding plot, with Economy and Science and Technology performing slightly at AUC = 0.53 and 0.55, respectively. With AUCs of 0.42 and 0.42, politics and education two subjects that were formerly among the weakest classes remain difficult. In many instances, the curves' closeness to the diagonal line indicates a performance that is almost identical to random guessing, indicating that additional optimization is required for model generalization. (iii) Right ROC Plot ( Figure 5 sub Figure (c) ): Health shows the greatest AUC of 0.58, followed by Culture and Tourism (AUC = 0.56) and Environmental (AUC = 0.55) in the third ROC plot, which shows modest increases in certain classes. These indicate that class discrimination is more successful. In the meantime, the AUCs for sport and education remain lower at 0.45 and 0.44, respectively. The classifier's ability to distinguish between classes is still not at its best, as indicated by the very small AUC range across every class. In general Values mostly cluster around 0.50, indicating a consistent pattern of moderate to low AUC scores across all three models. This indicates that most categories don't demonstrate high selective performance. Higher AUC scores across plots are typically obtained by classes like Culture and Tourism, Health, and Environmental, showing that these domains have more unique feature patterns. AUC values, however, are typically lower for categories like sport, foreign affairs, and education, indicating either a lack of feature representation or class overlap in the feature space. The sensitivity of the classification performance to the underlying model architecture or feature extraction method is shown by the variation in ROC curves and AUC values among the various models. In order to enhance performance in low-AUC categories, our results highlight the necessity for additional feature engineering improvement, model modification, or the integration of domain-specific knowledge. The accuracy of the CNN + Self-Attention model, illustrated in Figure 6 subfigure (a), increases gradually throughout the first few epochs before plateauing at around epoch 6, when it reaches a validation accuracy of roughly 0.90. The model hits an optimal point early in training, as evidenced by the validation loss flattening after epoch 4 but the training loss gradually declining. Yet, the small variation between the training and validation curves raises the possibility of minor overfitting, particularly in light of the lack of recurrent layers, which could make it more difficult for the model to accurately represent temporal dependencies in the input. The CNN + BiGRU + Self-Attention model, on the other hand, exhibits significantly faster convergence, with both training and validation accuracy above 0.90 during the first three epochs, as shown in Figure 6 subfigure (b). Indicating that the model learns rapidly and generalizes effectively, the training and validation losses decrease dramatically at the beginning and level off at epoch 4. The reduced difference between the accuracy and loss training and validation curves indicates that the BiGRU part improves the ability of the model to handle temporal information while maintaining the model's relative light weight and efficiency. A comparable trend can be noticed in the CNN + BiLSTM + Self-Attention model, which is shown in Figure 6 subfigure (c). It achieves high accuracy values above 0.95 in a short period of time and has the lowest training and validation losses of the three models. High generalization is showed by the fact that its validation performance constantly approaches training performance. The focus mechanism of self-attention and BiLSTM's capability to identify long-range dependencies in both directions work well together in this paradigm. It provides the best convergence speed and stability, despite being somewhat more computationally complex than BiGRU. To assess the interpretability and decision rationale of the classification model, LIME (Local Interpretable Model-Agnostic Explanations) was employed to visualize the most influential words contributing to predictions within the Health category. Figure 7 presents three representative examples where the model correctly classified input texts as health-related with high confidence.The model assigned a prediction probability of 0.88 for the Health class, with minor probabilities assigned to Science and Technology (0.04), Economy (0.02), and Environmental (0.02). The highlighted words such as “ሕክምና” (treatment), “እንክብካቤ” (care), and “የጤና” (of health) illustrate the model’s focus on semantically relevant tokens, reinforcing the correctness of the prediction. In Figure 8 the model's prediction for Health increased to 0.98, with negligible competition from other classes. The explanation highlights terms like “በሽታ” (disease), “ክስተት” (crisis), and “አካል” (body), all of which are core to the health domain. This near-perfect confidence suggests the model has developed a strong internal representation of medical and wellness-related language. Even with a slightly lower confidence level (0.93), Figure 9 indicates that Health still has a significant advantage over rival classes like Politics (0.05) and Foreign Affairs (0.01). Consistent health-related phrases like “የጤና”, “ሕክምና”, and “እንክብካቤ”, are highlighted by the model once more, indicating thematic consistency and robustness across samples. The model's predictions are based on significant linguistic features rather than meaningless word patterns, as the LIME visualizations generally reflect. The model has successfully internalized crucial semantic structures related to the health category, as seen by the periodic presence of domain-specific terms. 5. Conclusions A significant achievement in the field of explainable AI (XAI) and natural language processing (NLP) of underrepresented languages is the building and use of an attention-based hybrid deep learning model for multi-class Amharic news classification. Utilizing the blending of various neural network components and attention interprets that improve performance and interpretability, this model successfully tackles the particular linguistic challenges of the Amharic language, including complicated morphology, agglutinative structure, and sparse annotated data. A hybrid neural network, which usually integrates Convolutional Neural Networks (CNNs) along with Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Units (BiGRU), occupies the core of the architecture. Whereas the BiLSTM/BiGRU layers are adept at modeling long-range relationships in text sequences, the CNN layers are outstanding at extracting local and position-invariant information. The model's ability to selectively focus on semantically relevant words and phrases is further enhanced by the integration of a self-attention mechanism. Self-attention enables the model to offer more reliable and contextually aware representations of the input by giving relevant tokens more weight, regardless of where they are in the text. Notably, the system's transparency and reliability are improved by integrating Explainable AI (XAI) tools, particularly LIME (Local Interpretable Model-agnostic Explanations). This becomes especially essential in critical or high-stakes fields like the classification of legal or political news, where user trust and ethical AI adoption depend on explainability. In short, this attention-based hybrid model is scalable and interpretable, which makes it suitable for other low-resource languages and domains in addition to achieving state-of-the-art performance in multi-class Amharic news classification. It opens the door for more accountable, intelligible, and efficient NLP solutions for Amharic and beyond by integrating explainability frameworks with powerful deep learning architectures. Declarations Funding: Authors declare no funding for this research Competing interests: The authors declare that there is no competing interest Conflicts of interest: The authors declare that there is no conflict of interest Clinical trial: not applicable Ethics, Consent to Participate, and Consent to Publish declarations: not applicable. Availability of data: The data generated during and/or analysed the current study are not publicly available but are available from the corresponding author on reasonable request. Author contribution: Amlakie Aschale Alemu: Data curation, Visualization, Validation, Methodology, Writing, Reviewing and Editing. Malefia Demilie Melesse: Visualization, Methodology, Reviewing and Editing. Daniel Arega Mengesha : Conceptualization, Visualization, and document correction. Misganaw Aguate Widneh: Conceptualization, Formal analysis, Investigation References D. E. Id, G. Haile, and W. Taye, “recurrent unit model for Amharic next word prediction,” pp. 1–10, 2022, doi: 10.1371/journal.pone.0273156. B. M. Hailu, “Semantic Role Labeling for Amharic Text Using Multiple Embeddings and Deep Neural Network,” IEEE Access , vol. 11, no. March, pp. 33274–33295, 2023, doi: 10.1109/ACCESS.2023.3263147. N. A. Asfaw, B. H. Belay, and K. M. Alemu, “OPEN A deep learning approach for line-level Amharic Braille image recognition,” pp. 1–14, 2024. D. Endalie and G. Haile, “Automated Amharic News Categorization Using Deep,” vol. 2021, 2021, doi: 10.1155/2021/3774607. C. Science, C. Science, C. Science, and C. Science, “Explainable Machine Learning For Text Classification : A Novel Approach To Transparency And Interpretability,” vol. 103, no. 3, pp. 907–915, 2025. S. Gao, B. He, and F. Schilder, “On the Effectiveness of Pre-Trained Language Models for Legal Natural Language Processing : An Empirical Study,” IEEE Access , vol. 10, no. May, pp. 75835–75858, 2022, doi: 10.1109/ACCESS.2022.3190408. A. A. Fenta and S. Gebeyehu, “Automatic Idiom Identification Model for Amharic,” vol. 22, no. 8, 2023, doi: 10.1145/3606864. M. Degu, A. Olalekan, G. Kassa, and T. Bezabh, “Contextual word disambiguates of Ge ’ ez language with homophonic using machine learning,” Ampersand , vol. 12, no. December 2023, p. 100169, 2024, doi: 10.1016/j.amper.2024.100169. B. H. Belay, G. B. Gebremeskel, B. B. Bezabih, and S. Gebeyehu, “Factorized Recurrent Neural Network with Attention for,” vol. 22, no. 12, 2023, doi: 10.1145/3630607. E. F. Dinsa, M. Das, T. Urgessa, and A. Krishnaraj, “Automatic categorization of medical documents in Afaan Oromo using ensemble machine learning techniques,” Discov. Appl. Sci. , 2024, doi: 10.1007/s42452-024-06307-0. K. Wang, Y. Ding, and S. C. Han, Graph neural networks for text classification : a survey , vol. 57, no. 8. Springer Netherlands, 2024. doi: 10.1007/s10462-024-10808-0. D. Voskergian, R. Jayousi, and M. Yousef, “Topic selection for text classification using ensemble topic modeling with grouping , scoring , and modeling approach,” pp. 1–19, 2024. T. Thaminkaew, P. Lertvittayakumjorn, and P. Vateekul, “Prompt-Based Label-Aware Framework for Few-Shot Multi-Label Text Classification,” IEEE Access , vol. 12, no. January, pp. 28310–28322, 2024, doi: 10.1109/ACCESS.2024.3367994. K. Taha, P. D. Yoo, C. Yeun, D. Homouz, and A. Taha, “A comprehensive survey of text classification techniques and their research applications : Observational and experimental insights,” Comput. Sci. Rev. , vol. 54, no. July, p. 100664, 2024, doi: 10.1016/j.cosrev.2024.100664. H. Sakai and S. S. Lam, “Large Language Models For Health Care Text Classification : A Systematic Review arXiv : 2503 . 01159v1 [ cs . CL ] 3 Mar 2025”. T. Kuzman, “LLM Teacher-Student Framework for Text Classification With No Manually Annotated Data : A Case Study in IPTC News Topic Classification,” vol. 13, no. January, pp. 35621–35633, 2025, doi: 10.1109/ACCESS.2025.3544814. A. Jahin, S. H. Shovon, M. F. Mridha, and R. Islam, “OPEN A hybrid transformer and attention based recurrent neural network for robust and interpretable sentiment analysis of tweets,” pp. 1–25, 2024. S. Hirpassa and G. S. Lehal, “Heliyon Improving part-of-speech tagging in Amharic language using deep neural network,” Heliyon , vol. 9, no. 7, p. e17175, 2023, doi: 10.1016/j.heliyon.2023.e17175. S. Gite, H. Khatavkar, K. Kotecha, and S. Srivastava, “Explainable stock prices prediction from financial news articles using sentiment analysis,” pp. 1–21, 2021, doi: 10.7717/peerj-cs.340. A. Nasayreh et al. , “Journal of King Saud University - Computer and Jordanian banknote data recognition : A CNN-based approach with attention mechanism,” J. King Saud Univ. - Comput. Inf. Sci. , vol. 36, no. 4, p. 102038, 2024, doi: 10.1016/j.jksuci.2024.102038. S. Khan et al. , “BiCHAT : BiLSTM with deep CNN and hierarchical attention for hate speech detection,” J. King Saud Univ. - Comput. Inf. Sci. , vol. 34, no. 7, pp. 4335–4344, 2022, doi: 10.1016/j.jksuci.2022.05.006. Z. Dai and Y. He, “A non-functional requirements classification model based on cooperative attention mechanism fused with label embedding,” Comput. Electr. Eng. , vol. 121, no. September 2024, p. 109856, 2025, doi: 10.1016/j.compeleceng.2024.109856. C. Rudin, “Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,” Nat. Mach. Intell. , vol. 1, no. 5, pp. 206–215, 2019, doi: 10.1038/s42256-019-0048-x. J. R. Jim, M. A. R. Talukder, P. Malakar, M. M. Kabir, K. Nur, and M. F. Mridha, “Recent advancements and challenges of NLP-based sentiment analysis: A state-of-the-art review,” Nat. Lang. Process. J. , vol. 6, no. January, p. 100059, 2024, doi: 10.1016/j.nlp.2024.100059. V. Buhrmester, D. Münch, and M. Arens, “Analysis of Explainers of Black Box Deep Neural Networks for Computer Vision: A Survey,” Mach. Learn. Knowl. Extr. , vol. 3, no. 4, pp. 966–989, 2021, doi: 10.3390/make3040048. Q. Xu et al. , “Applications of Explainable AI in Natural Language Processing,” Glob. Acad. Front. , vol. 2, no. 3, pp. 51–64, 2024. S. Luo, H. Ivison, S. C. Han, and J. Poon, “Local Interpretations for Explainable Natural Language Processing: A Survey,” ACM Comput. Surv. , vol. 56, no. 9, 2024, doi: 10.1145/3649450. E. Hashmi, S. Y. Yayilgan, M. M. Yamin, S. Ali, and M. Abomhara, “Advancing Fake News Detection : Hybrid Deep Learning With FastText and Explainable AI,” IEEE Access , vol. 12, no. March, pp. 44462–44480, 2024, doi: 10.1109/ACCESS.2024.3381038. S. M. Carta, S. Consoli, L. Piras, A. S. Podda, and D. R. Recupero, “Explainable Machine Learning Exploiting News and Domain-Specific Lexicon for Stock Market Forecasting,” IEEE Access , vol. 9, pp. 30193–30205, 2021, doi: 10.1109/ACCESS.2021.3059960. M. M. Azad and H. S. Kim, “An explainable artificial intelligence-based approach for reliable damage detection in polymer composite structures using deep learning,” Polym. Compos. , no. September 2024, pp. 1536–1551, 2024, doi: 10.1002/pc.29055. I. Of, D. Learning, M. In, and N. Language, “Article Info,” vol. 3, no. 2, pp. 56–66, 2025. M. S. H. Apu, M. S. Islam, and T. T. Aurpa, “Explainable AI for Sentiment Analysis of Human Metapneumovirus (HMPV) Using XLNet,” 2025, [Online]. Available: http://arxiv.org/abs/2502.01663 N. Pospelova, A. Tarasova, N. Subbotina, N. Koroleva, N. Raimova, and E. L. Lydia, “Explainable Artificial Intelligence and Natural Language Processing for Unraveling Deceptive Contents,” Fusion Pract. Appl. , vol. 14, no. 2, pp. 146–158, 2024, doi: 10.54216/FPA.140212. K. Zahoor, N. Z. Bawany, and T. Qamar, “Evaluating text classification with explainable artificial intelligence,” IAES Int. J. Artif. Intell. , vol. 13, no. 1, pp. 278–286, 2024, doi: 10.11591/ijai.v13.i1.pp278-286. T. Bao, N. Ren, R. Luo, B. Wang, G. Shen, and T. Guo, “A BERT-Based Hybrid Short Text Classification Model Incorporating CNN and Attention-Based BiGRU,” J. Organ. End User Comput. , vol. 33, no. 6, pp. 1–21, 2021, doi: 10.4018/joeuc.294580. M. Ahmed, M. S. Hossain, R. Ul Islam, and K. Andersson, “Explainable Text Classification Model for COVID-19 Fake News Detection,” J. Internet Serv. Inf. Secur. , vol. 12, no. 2, pp. 51–69, 2022, doi: 10.22667/JISIS.2022.05.31.051. A. Caleb, “Multi-Class Fake News Detection Using Deep Learning and Explainable AI : A Transformer-Gradient Boosting Approach,” no. March, 2025. M. Cesarini, L. Malandri, F. Pallucchini, A. Seveso, and F. Xing, “Explainable AI for Text Classification: Lessons from a Comprehensive Evaluation of Post Hoc Methods,” Cognit. Comput. , pp. 3077–3095, 2024, doi: 10.1007/s12559-024-10325-w. B. Jang, M. Kim, G. Harerimana, S. Kang, and J. W. Kim, “applied sciences Bi-LSTM Model to Increase Accuracy in Text Classification : Combining Word2vec CNN and Attention Mechanism,” 2020. J. Blake, A. B. U. Saleh, M. Miah, K. Kredens, J. Shin, and S. Member, “Detection of AI-Generated Texts : A Bi-LSTM and Attention-Based Approach,” IEEE Access , vol. 13, no. April, pp. 71563–71576, 2025, doi: 10.1109/ACCESS.2025.3562750. M. Yibeyin, Y. Tehone, A. Liyih, and M. Fentie, “Public opinion mining in social media about Ethiopian broadcasts using deep learning,” pp. 1–10, 2024. M. M. Hasan et al. , “Hybrid deep learning: a comparative study on ai algorithms in natural language processing for text classification,” Bull. Electr. Eng. Informatics , vol. 14, no. 1, pp. 551–559, 2025, doi: 10.11591/eei.v14i1.7617. S. Murindanyi, B. A. Yiiki, A. Katumba, and J. Nakatumba-Nabende, “Explainable Machine Learning Models for Swahili News Classification,” ACM Int. Conf. Proceeding Ser. , pp. 12–18, 2023, doi: 10.1145/3639233.3639250. D. Rohera et al. , “A Taxonomy of Fake News Classification Techniques: Survey and Implementation Aspects,” IEEE Access , vol. 10, pp. 30367–30394, 2022, doi: 10.1109/ACCESS.2022.3159651. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 24 Jun, 2025 Reviews received at journal 19 Jun, 2025 Reviewers agreed at journal 19 Jun, 2025 Reviews received at journal 16 Jun, 2025 Reviewers agreed at journal 10 Jun, 2025 Reviewers agreed at journal 10 Jun, 2025 Reviewers agreed at journal 10 Jun, 2025 Reviewers invited by journal 06 Jun, 2025 Editor assigned by journal 03 Jun, 2025 Submission checks completed at journal 03 Jun, 2025 First submitted to journal 27 May, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6759877","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":468114032,"identity":"33faeb91-4967-4ee8-92f1-f1bbe8a9667c","order_by":0,"name":"Amlakie Aschale Alemu","email":"data:image/png;base64,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","orcid":"","institution":"Debre Tabor University","correspondingAuthor":true,"prefix":"","firstName":"Amlakie","middleName":"Aschale","lastName":"Alemu","suffix":""},{"id":468114033,"identity":"a1d7195f-5559-4c0f-8a5c-efbf65d2091e","order_by":1,"name":"Malefia Demilie Melese","email":"","orcid":"","institution":"Debre Tabor University","correspondingAuthor":false,"prefix":"","firstName":"Malefia","middleName":"Demilie","lastName":"Melese","suffix":""},{"id":468114034,"identity":"fad61935-5ecc-4159-9a8a-d312a2baf8ae","order_by":2,"name":"Daniel Arega Mengesha","email":"","orcid":"","institution":"Debre Tabor University","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"Arega","lastName":"Mengesha","suffix":""},{"id":468114035,"identity":"21cd38f2-92d2-4390-a362-82fd799d9549","order_by":3,"name":"Misganaw Aguate Widneh","email":"","orcid":"","institution":"Debre Tabor University","correspondingAuthor":false,"prefix":"","firstName":"Misganaw","middleName":"Aguate","lastName":"Widneh","suffix":""}],"badges":[],"createdAt":"2025-05-27 13:23:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6759877/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6759877/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":84240300,"identity":"e7d1fe05-bf2e-4338-888e-227f01211a17","added_by":"auto","created_at":"2025-06-09 15:42:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":287926,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOverall system architecture of the study\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/abe5da22f0cc4d2d3ba7281a.png"},{"id":84240632,"identity":"de267478-fa83-4a8a-a7f4-b203ff37da40","added_by":"auto","created_at":"2025-06-09 15:50:58","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":270630,"visible":true,"origin":"","legend":"\u003cp\u003eProposed architeture of CNN + BiLSTM with Attention Mechanism\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/72aa18ffff64576a623f9f72.png"},{"id":84240633,"identity":"953421d4-27ee-414d-ba10-af9877bec040","added_by":"auto","created_at":"2025-06-09 15:50:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":134559,"visible":true,"origin":"","legend":"\u003cp\u003eModel performance comparisons with Attention Mechanism\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/21ba7620eff2fe1c7a7c4bf7.png"},{"id":84240309,"identity":"e407bede-10a7-48a9-a415-e74063fe8c0d","added_by":"auto","created_at":"2025-06-09 15:42:58","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":267972,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion Matrices of Attention Mechanism with three hybrid deep learning algorithms:\u003c/p\u003e\n\u003cp\u003e(a) CNN + BiLSTM + Self-Attention, (b) CNN+BiGRU + Self-Attention, (c) CNN+Self-Attention\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/1d0a80d9299a0e4d7e80697c.png"},{"id":84240634,"identity":"d303cd2b-f8f4-441a-99e1-a3bcf93f7492","added_by":"auto","created_at":"2025-06-09 15:50:58","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":382241,"visible":true,"origin":"","legend":"\u003cp\u003eROC Curve Analysis for Multi-class News Categorization: (a) CNN + BiLSTM + Self-Attention, (b) CNN+BiGRU + Self-Attention, (c) CNN+Self-Attention\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/9fdd56239e07ec03fc887cbf.png"},{"id":84240639,"identity":"e3d44811-7572-4e5e-9165-5b88ddf79f1b","added_by":"auto","created_at":"2025-06-09 15:50:58","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":322256,"visible":true,"origin":"","legend":"\u003cp\u003eAccuracy and Loss versus epoch plot: (a) for CNN + BiGRU + Self-Attention, (b) for CNN + BiLSTM + Self-Attention and (c) for CNN + Self-Attention\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/6123e8772b00dd25512b7dad.png"},{"id":84240312,"identity":"11cf8dec-095b-4e13-b6c6-7aec0d0bad3f","added_by":"auto","created_at":"2025-06-09 15:42:58","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":100825,"visible":true,"origin":"","legend":"\u003cp\u003eGraphical representation of model interpretability for CNN using LIME\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/57bc1af209a8c5d41b293742.png"},{"id":84240322,"identity":"b54c7de6-4ffb-4be3-8757-7bcd2155c5fc","added_by":"auto","created_at":"2025-06-09 15:42:59","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":156612,"visible":true,"origin":"","legend":"\u003cp\u003eGraphical representation of model interpretability for CNN+BiLSTM using LIME\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/41e2495f6c0cf64e76dc2c02.png"},{"id":84240636,"identity":"2d18d9f9-2280-4dcd-86eb-9361000e423b","added_by":"auto","created_at":"2025-06-09 15:50:58","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":102367,"visible":true,"origin":"","legend":"\u003cp\u003eGraphical representation of model interpretability for CNN+BiGRU using LIME\u003c/p\u003e","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/bfd53dbc8cec93253cb84d5f.png"},{"id":84242582,"identity":"42d52069-7863-4d86-9750-06ae7344bc5a","added_by":"auto","created_at":"2025-06-09 16:09:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2989754,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6759877/v1/aa0bf6aa-a9ed-4001-8797-6494680aa1bf.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Attention Based Hybrid Deep Learning models for Multi class Amharic News Categorization with Explainable AI","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eLow-resource languages like Amharic, which is one of the most frequently spoken languages in Africa and Ethiopia\u0026apos;s official working language, have also been affected by the exponential growth in digital content worldwide [1], [2], [3],[4]. Automatic classification of this data is essential for sentiment analysis, information retrieval, focused on news delivery, and effective information framework [5]. However, because of its rich morphology, complicated syntactic structure, and absence of annotated corpora and computational resources, Amharic presents particular difficulties for natural language processing (NLP) [6], [7], [8], [9]. Support Vector Machines (SVM) and Na\u0026iuml;ve Bayes are two examples of traditional machine learning techniques for text categorization that mostly rely on human feature engineering and shallow representations like bag-of-words or TF-IDF, which frequently fall short in capturing the semantic and contextual richness of texts [10]. Whereas end-to-end learning of hierarchical features made possible by deep learning models has transformed natural language processing (NLP) [6], [9], acceptance of them in Amharic and other low-resource languages is still quite restricted [9], [10].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhen attempting to do multi-class classification [11], [12], [13], [14], [15], [16] of Amharic articles, we present a hybrid deep learning architecture that employs Convolutional Neural Networks (CNNs) with Bidirectional GRU (CNN + BiGRU)\u0026nbsp;and\u0026nbsp;Bidirectional LSTM (CNN + BiLSTM) networks, enriched with a self-attention mechanism. While BiLSTM layers model long-range dependencies and contextual flow in both forward and backward directions, CNN layers are used to extract local n-gram features that capture syntactic patterns [17], [18], [19]. The model can concentrate on contextually relevant tokens that are required for classification because the attention mechanism dynamically weights the importance of various words in a sequence [15], [20], [21], [22]. Furthermore, one of the most major drawbacks of deep learning models is their black-box dimension, which makes them difficult to interpret and undermines user confidence, particularly in low-resource environments where linguistic proficiency is crucial [23], [24], [25]. We use Explainable AI (XAI) techniques in our framework to tackle this. While techniques such as Local Interpretable Model-Agnostic Explanations (LIME) are used to provide after the fact model explanations, attention visualization specifically emphasizes the role of individual tokens in the final prediction [26], [27], [28].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA newly developed or designed multi-class Amharic news dataset with annotations across a variety of topical groups is used to test the model that we propose. According to experimental evaluations, our hybrid architecture performs better in terms of classification accuracy, macro F1-score, and robustness than both standalone deep learning models and conventional baselines. Likewise the model\u0026apos;s ability to offer meaningful and intelligible insights into the categorization reasoning is verified by attention corresponds and LIME-based explanations [29], [30]. By (i) presenting an efficient deep learning pipeline designed for the Amharic language, (ii) resolving the interpretability issue with integrated XAI techniques, and (iii) offering a useful standard for further research on Amharic news classification and multilingual explainable NLP systems, this study makes a major contribution to the field of NLP for low-resource languages. The primary contributions of the study are as follows.\u003c/p\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n \u003cli\u003eWe propose a hybrid model that combines Convolutional Neural Networks (CNN) with Bidirectional GRU (CNN + BiGRU)\u0026nbsp;and\u0026nbsp;Bidirectional LSTM (CNN + BiLSTM)\u0026nbsp;networks, effectively capturing both local semantic patterns and long-range dependencies in Amharic news texts. This architecture is further enhanced by a self-attention mechanism, allowing the model to selectively focus on the most informative words in a sequence.\u003c/li\u003e\n \u003cli\u003eThe model is specifically designed and optimized for the Amharic language, addressing its unique linguistic characteristics such as rich morphology, syntactic variations, and limited annotated datasets. This research bridges a critical gap in deep learning applications for low-resource languages.\u003c/li\u003e\n \u003cli\u003eWe integrate Explainable AI (XAI) techniques into the classification framework. Local Interpretable Model-Agnostic Explanations (LIME) offers post-hoc explanations. Together, these methods enhance transparency and user trust in the system\u0026rsquo;s decisions.\u003c/li\u003e\n \u003cli\u003eA multi-class labeled dataset of Amharic news articles covering categories such as politics, economy, health, sports, and technology is compiled and annotated. This dataset contributes to the Amharic NLP community and serves as a benchmark for future research.\u003c/li\u003e\n \u003cli\u003eThis work contributes to the broader effort of democratizing NLP technologies by advancing research in low-resource African languages. It demonstrates that deep learning models, when properly adapted and explained, can perform competitively even in resource-scarce contexts.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"2. Related Works","content":"\u003cp\u003eThe paper in [5] provides an in-depth examination of XML approaches used in text classification models, including post-hoc interpretability frameworks such as LIME and SHAP, attention mechanisms, and feature significance strategies. Additionally, by balancing the trade-offs between interpretability and performance, they evaluate these methods\u0026apos; ability to improve model transparency. The study also explores how XML may develop in text classification in the future, including how user-focused explanations and legal requirements for AI transparency could get incorporated. According to the findings, explainability helps to detect model biases, improve overall performance, and increase the trustworthiness of the model. The work [28] provides an integrated strategy for fake news detection, applying three publicly available datasets: WELFake, FakeNewsNet, and FakeNews Prediction. They merged FastText word embedding with various Machine Learning and Deep Learning methods, further refining these algorithms with regularization and hyperparameter optimization to prevent overfitting and enhance model generalization. In addition, they utilized state-of-the-art transformer-based models like as BERT, XLNet, and RoBERTa, improving them through hyperparameter adjustments. In the concluding phase, explainable AI modeling was applied using Local Interpretable Model-Agnostic Explanations, and Latent Dirichlet Allocation to get deeper insights into the model\u0026rsquo;s decision-making process.\u003c/p\u003e\n\u003cp\u003eThe work [31] explores the integration of Explainable Artificial Intelligence (XAI) methodologies. This study\u0026apos;s main goal is to assess how various XAI techniques can be used to interpret and explain the choices made by NLP-based misinformation classifiers. To investigate model behavior and feature importance, XAI techniques such as SHAP (SHapley Additive Explanations), LIME (Local Interpretable Model-agnostic Explanations), and attention visualization were used. The results show that XAI techniques greatly increase transparency by highlighting key words and phrases that influence model decisions. \u0026nbsp;The study [32] explores how sentiment analysis, utilizing data from social media platforms like YouTube, can improve thier comprehension of how the public responds to HMPV. It emphasizes how crucial it is to monitor public state of mind, which ranges from fear to trust, in order to inform regulations, address false information, guide health communications, and promote adherence to preventive measures during outbreaks. In sentiment classification tasks, the study uses sophisticated transformer models, especially XLNet, and achieves an accuracy of 93.50%. Furthermore, they apply SHAP to integrate explainable AI (XAI) to make it transparent how the model determines the main determinants of public perception.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn study [33] provides a new deep learning-based method for detecting misleading content on social media using the sine cosine algorithm (SCADL-DCDSM). To classify emotions, the SCADL-DCDSM technique uses a hyper-parameter tuning strategy combined with the ensemble learning process. Furthermore, three models long short-term memory (LSTM), extreme learning machine (ELM), and attention-based recurrent neural network (ARNN) are used in the SCADL-DCDSM technique for sentiment classification. Lastly, SCA can be used to improve the DL models\u0026apos; hyper-parameter selection. In order to improve accurate deceptive content recognition, the SCADL-DCDSM system incorporates the explainable artificial intelligence (XAI) system LIME, which has been used for an extensive explainability and grasp of the black-box process. The results of the model showed that, when considering several metrics, the SCADL-DCDSM methodology produces the best result. \u0026nbsp;According to study [34] human beings are able to evaluate machine learning models for accuracy, fairness, and dependability due to the developing discipline of explainable artificial intelligence (XAI). Using a well-known Explainable AI method called local interpretable model agnostic explanations (LIME), they build on our earlier research and conduct a thorough investigation of the model developed for text categorization and sentiment analysis. The findings confirm that explainable AI tools are necessary for evaluating machine learning models because accuracy and other associated metrics do not guarantee the model\u0026apos;s correctness, fairness, and dependability. Furthermore, they use LIME to compare the interpretability and explainability of different machine learning techniques.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA feature fusion framework based on the bidirectional encoder representations from transformers (BERT) is proposed in paper [35]. This hybrid approach trains the word vector representation using BERT. Static characteristics are extracted by convolutional neural networks (CNNs). Furthermore, to gather contextual data, a bi-gated recurrent neural network (BiGRU) is used. In addition, an approach for determining the weight for important words is introduced. The proposed approach performs noticeably better than the other advanced baseline techniques, according to the experimental data. In the paper [36] they \u0026nbsp;presented a hybrid LIME-BiLSTM model in which LIME ensures explainability and transparency while BiLSTM provides classification accuracy. The proposed framework becomes understandable in this integrated model since LIME explains the prediction and performs similarly to the original model. Kendall\u0026apos;s tau correlation coefficient is used to assess this model\u0026apos;s explainability performance. likewise they \u0026nbsp;use a number of machine learning models and compare how well they work. Because the model adopts an integrated approach, they examined and contrasted the computational overhead of our suggested model with the other approaches.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe study in [37] offers a multi-class approach to detecting fake news that combines Gradient Boosting methods with Transformer-based deep learning models. Furthermore, interpretability and insights into model predictions are provided by Explainable AI (XAI) techniques as SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations). The experimental findings, which were obtained using benchmark datasets like LIAR and FakeNewsNet, show that the proposed mixed approach performs better than standalone models in terms of interpretability, accuracy, and F1-score. A major discrepancy in the evaluation of eXplainable Artificial Intelligence (XAI) techniques for text categorization is addressed in the research [38]. While current frameworks concentrate on evaluating XAI in domains like visual analytics and recommender systems, an extensive evaluation is missing. Recent afterwards XAI techniques are surveyed and categorized in our study based on their output format and explanatory scope. The main conclusions show that feature-based explanations are more accurate than rule-based ones. These revelations improve knowledge of XAI in text classification and provide insightful advice for creating successful XAI systems.\u0026nbsp;Table 1 shows how our study\u0026apos;s model comparison utilizing accuracy is depicted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;Table 1:\u003c/strong\u003e Comparison of the proposed model with existing state of the art models\u003c/p\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eArticle References\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTasks\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eApplying \u0026nbsp; \u0026nbsp; Attention Mechanism\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eApplying Model\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eInterpretability\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[4]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eCategorization\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e93.79%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[39]\u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eClassification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eCNN+BiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e91.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[40]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eClassification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eBiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e88.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[41]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eClassification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eCNN+BiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e96.08%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[42]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eClassification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eBiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e96.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eClassification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e92.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e[44]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eClassification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003eCNN+BiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOur study\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCategorization\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 114px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCNN\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eCNN+BiGRU\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eCNN+BiLSTM\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e90\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e92.7\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e97\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eYes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 130px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eYes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e"},{"header":"3. Methods and Materials","content":"\u003cp\u003eThe following figure 1 illustrates a comprehensive end-to-end pipeline for Amharic text news classification, starting from data acquisition to model interpretation. The proposed approach for multi-class Amharic news classification follows a structured pipeline that integrates data preparation, hybrid deep learning model construction, and explainable AI mechanisms.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.1.\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003eAmharic Text News Data Collection:\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThroughout this stage, the raw Amharic articles are collected from various digital sources and online news sites. To ensure the diversity and the representativeness of the categorization task, the data covers a wide range of areas, including agriculture, culture and tourism, education, economy, environment, foreign affairs, health, politics, science and technology, and sports. Since there doesn\u0026apos;t appear to a publicly available Amharic document classification corpus, we collected 1200 Amharic news documents from ten major news categories, each news category containing 1500 sentences, in order to train and test the deep learning model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.2.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eData Cleaning and Preprocessing:\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003eNoise, inconsistencies, and unnecessary characters like punctuation, stop-words, numbers, and symbols are frequently found in collected text data. This process involves the following steps: (1) Unicode standardization and text normalization; (2) removal of non-Amharic tokens, special characters, and redundant white spaces; and (3) normalizing of common morphological variants to ensure consistency. After the fact that the data is cleaned and preprocessed, which includes normalizing the text for consistency and removing noise like punctuation, numbers and non-Amharic characters. To get the text prepared for analysis, this step may include lemmatization of stemming, and stop word removal.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e3.3. Data Labeling and Tokenization:\u0026nbsp;\u003c/strong\u003eData labeling and tokenization come next after preprocessing. In order to get ready the text for machine learning input, each article is given a label that corresponds to its category (politics, sports, business, etc.). The text is then tokenized and divided into smaller chunks, such as words or sub-words. \u0026nbsp;Every news item in this stage is given the appropriate class label. Splitting texts into linguistically meaningful units (tokens), usually words or sub-words, is known as tokenization. To transform textual tokens into dense vector representations appropriate for deep learning models, word embedding tools like FastText are optionally used here.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e3.4.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eDataset Splitting:\u0026nbsp;\u003c/strong\u003eTo ensure equal representation across categories, the dataset is divided into training and testing sets. This stage makes it easier to learn the model and evaluate it accurately. Training and testing are done in a standard ratio, like 80:20. The labeled dataset is subsequently split into subsets for testing and training, usually known as dataset splitting. We employed a total of 2200 datasets, of which 1760 documents were used for training and the remaining 440 documents for testing.\u0026nbsp;This makes it possible to test the model\u0026apos;s generalization performance on one set of data and train it on another.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.5.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eModel Building (Hybrid Deep Learning with Self-Attention):\u0026nbsp;\u003c/strong\u003eDeep learning architectures are used in the pipeline\u0026apos;s basic process of building models. CNN (Convolutional Neural Network), CNN blended with BiGRU (Bidirectional Gated Recurrent Unit), and CNN integrated with BiLSTM (Bidirectional Long Short-Term Memory) are among the models that have been highlighted. In order to better comprehend contextual meaning in Amharic text, these models are further improved with self-attention mechanisms that enable them to concentrate on the most pertinent portions of the text independent of word position. An entirely novel hybrid architecture develops that includes:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eCNN (Convolutional Neural Networks)\u0026nbsp;to capture local spatial features and patterns in the text.\u003c/li\u003e\n \u003cli\u003eBidirectional GRU (CNN + BiGRU)\u0026nbsp;and\u0026nbsp;Bidirectional LSTM (CNN + BiLSTM)\u0026nbsp;to model long-range and bidirectional dependencies in sequences.\u003c/li\u003e\n \u003cli\u003eA self-attention mechanism is integrated to weigh the contextual importance of tokens, allowing the model to focus on semantically relevant parts of the text. This hybrid structure combines the strengths of both convolutional and recurrent layers, enhanced by attention to improve classification performance and interpretability.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eA hybrid deep learning model which includes Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM) networks, and a Self-Attention mechanism for text classification is depicted in the architecture figure 2.\u0026nbsp;There are multiple logical sections to this model\u0026apos;s structure: A preprocessed text dataset that has undergone common cleaning techniques including tokenization, stop-word removal, and normalization is forwarded to the input phase. The word embedding section accepts this processed data and uses pre-trained FastText word embedding to turn each word into a dense vector. By using subword information to determine both the morphology and meaning of words, these embeddings enhance the representation. Beyond that, the embedded sequences get passed to the feature extraction section, which uses a number of 1D convolutional filters to identify local patterns in the text, like n-grams. By emphasizing grammatical structures, the convolutional layers serve as effective feature extractors. Following that, max-pooling layers increase efficiency and generalization by reducing the dimensionality of the feature maps while keeping the most noticeable features. Bidirectional LSTM (BiLSTM) networks are used in the sequence learning section to identify long-range dependencies in the text. The BiLSTM outputs are subsequently submitted to a self-attention layer via the attention mechanism section. By using this technique, the model can focus on informative terms and eliminate out irrelevant ones by allocating weights to various tokens according to their significance to the classification task. By attending to critical words across the entire sequence, the self-attention module enhances the interpretability and effectiveness of the model.\u003c/p\u003e\n\u003cp\u003eThe attention layer\u0026apos;s enhanced characteristics are subsequently processed via a dense layer with ReLU activation and a softmax output layer in the classification stage. The softmax layer provides class probabilities, which show the model\u0026apos;s prediction for every input sequence, while the ReLU layer offers non-linearity, allowing the network to learn intricate patterns. This CNN-BiLSTM-Self Attention model is especially well-suited for sophisticated natural language processing tasks including sentiment analysis, document categorization, and intent identification because it effectively determines local patterns, long-term dependencies, and global contextual relevance. During training, the model is represented as the trained model, which can now predict new, unseen text.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.6.\u003c/strong\u003e \u003cstrong\u003eModel Evaluation:\u0026nbsp;\u003c/strong\u003eThe next step involves model evaluation, where performance metrics such as accuracy, precision, recall, and F1-score are used to assess how well the model performs. The model is trained on the labeled training dataset using appropriate loss functions and optimization algorithms. Evaluation is conducted on the test set using standard classification metrics. In this work, we evaluate the proposed models\u0026apos; performance using evaluation measures like accuracy, precision, recall, and F1 score which are formulated as the following equations 1-4.\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" style=\"width: 763px; height: 243.942px;\" width=\"763\" height=\"243.942\"\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003eWhere, TP, TN, FP and FN represent true positive, true negative, false positive and false negative respectively. These metrics provide insight into the model\u0026apos;s effectiveness across all news categories.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.7 Model Interpretability with Explainable AI (XAI\u003c/strong\u003e\u003cstrong\u003e):\u003c/strong\u003e Subsequently, the pipeline blends model interpretability through the use of tools such as LIME (Local Interpretable Model-agnostic Explanations), which improve transparency and reliability by offering insights into the model\u0026apos;s decision-making process. LIME (Local Interpretable Model-Agnostic Explanations) is a subsequent interpretability tool that produces human-readable explanations for individual predictions in order to overcome the black-box feature of deep models and increase user trust. By ensuring transparency, this element assists in the understanding and validation of the system\u0026apos;s behavior by users and domain experts. All things taken into account, this process integrates preprocessing, advanced neural network models, and interpretability techniques to produce a reliable and understandable system, representing a comprehensive and contemporary approach to Amharic text classification. The proposed technique for classifying multi-class Amharic news uses a structured workflow which integrates the building of hybrid deep learning models with data preprocessing.\u003c/p\u003e"},{"header":"4. Results and Discussions","content":"\u003cp\u003eTo compare convolutional recurrent neural networks with distinct characteristics and assess the study\u0026apos;s performance in regard to the challenges we looked at, this section offers the built models and the experiments with the results collected. We must analyze the dataset we used to build our model in order to accomplish this. The outcomes produced by the implementation are investigated, and the metrics are used to assess the results. To find the one that would work best, we experimented with several combinations of the model\u0026apos;s hyper-parameters. Thus, as indicated in table 2,\u0026nbsp;we select the following hyper-parameters.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2: Hyper-parameter configuration details for the proposed model\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"6\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHyper-parameters\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eModels\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eEpoch\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eBatch size\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eLearning rate\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eOptimizer\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eActivation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDropout\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1e-5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eReLu\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSoftMax\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCNN+BiGRU\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1e-5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eReLu\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSoftMax\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCNN+BiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1e-5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eReLu\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSoftMax\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eGenerally, when looking at the whole hyper-parameters and models of the experiments, table 2 gives a short and precise analysis of the experimental result of this study for the selection of hyperparameters.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3:\u0026nbsp;Evaluation metrics of deep learning models Attention mechanism\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eEvaluation Metrics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eModels\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1_Score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e89.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e89.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e89.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCNN+BiGRU\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e92.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e92.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e92.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e92.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCNN+BiLSTM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e96.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eThree deep learning models CNN, CNN+BiGRU, and CNN+BiLSTM are compared in the evaluation metrics which is plotted in table 3 based on four important classification metrics: accuracy, precision, recall, and F1 score. The baseline model, the Convolutional Neural Network (CNN), achieves an overall accuracy of 90, an F1 score of 89.7, a precision of 89.8, and a recall of 89.9. We notice considerable improvements in every metric when a Bidirectional Gated Recurrent Unit (BiGRU) is integrated to the CNN. The accuracy, precision, recall, and F1 score of the CNN+BiGRU model are 92.2, 92.4, and 92.3, respectively. The integration of BiGRU, which enables the model to account for both past and future context in the input sequences a vital feature in tasks like natural language processing is responsible for this enhancement. The CNN+BiLSTM model provides the most notable performance. Out of the three models, this combination produces the highest accuracy (97), precision (96.7), recall (96.8), and F1 score (96.8). The enhanced outcomes indicate that BiLSTM is even more successful than BiGRU at figuring out long-term dependencies in the data. When everything is taken into account, the performance trends render it abundantly evident that blending CNN with recurrent architectures improves the model\u0026apos;s grasp of sequential or contextual patterns in data. CNN+BiLSTM consistently perform better than CNN+BiGRU among the two hybrid models, despite at a possibly better computational cost. CNN+BiLSTM were the proposed alternative for applications requiring the highest level of accuracy and robustness. Even though it is less precise, the standalone CNN model might nevertheless work well in contexts with limited resources or a lightweight.\u003c/p\u003e\n\u003cp\u003eThree deep learning architectures are compared in\u0026nbsp;Figure 3\u0026nbsp;based on four common classification metrics: accuracy, precision, recall, and F1 score. CNN with Self-Attention, CNN with BiGRU and Self-Attention, and CNN with BiLSTM and Self-Attention are among the models. With values nearly 97, the CNN+BiLSTM+Self-Attention model outperforms than the other two models in every criterion, indicating higher classification capabilities. According to this, introducing bidirectional LSTM layers greatly improves the model\u0026apos;s capacity to identify textual sequential dependencies, which are then further refined by the attention mechanism.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConsidering metric values ranging from 92 to 92.3, the CNN+BiGRU+Self-Attention model shows a moderate improvement over the CNN+Self-Attention baseline, indicating that GRU-based sequence modeling provides a useful improvement over convolutional layers alone. The CNN+Self-Attention model, on the other hand, performs comparatively poorly in grasping contextual complexities, even if it is still robust. Its scores are just below 90. The efficiency of integrating sequential neural layers with attention mechanisms to improve text classification tasks in complicated language datasets has been shown by a gradual rise in all four measures across model configurations.\u003c/p\u003e\n\u003cp\u003eTable 4: Classification Report for CNN+Self Attention\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"858\" height=\"348\"\u003e\u003c/p\u003e\n\u003cp\u003eWhen recurrent and attention mechanisms are integrated, the models\u0026apos; classification performance clearly shows an upward trend as indicated in table 4. With a 90% overall accuracy and weighted and macro averages of precision, recall, and F1-score of 0.90, the CNN+Self-Attention model offers a strong foundation. The model\u0026apos;s performance is marginally worse in the areas of Economy (F1-score = 0.79) and Politics (F1-score = 0.72), but it exceeds in Sport (F1-score = 0.99) and Science and Technology (F1-score = 0.95). These differences imply that CNN by itself does a good job of identifying patterns in some domains, especially when the terminology or textual structure is consistent, but it may have difficulties with more complex or context-heavy categories.\u003c/p\u003e\n\u003cp\u003eTable 5: Classification Report for CNN+BiGRU+Self Attention\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"868\" height=\"351\"\u003e\u003c/p\u003e\n\u003cp\u003eIn Table 5 the model\u0026apos;s contextual understanding is greatly enhanced by incorporating a BiGRU layer in along with CNN and Self-Attention, increasing overall accuracy to 92%. The precision, recall, and F1-score weighted and macro averages all rise to 0.92. In comparison to the former model, this model shows noticeable enhancements in Politics (F1 = 0.86) and Economy (F1 = 0.84), while it achieves almost perfect performance in classes like Education (F1 = 0.98) and Sport (F1 = 1.00). The BiGRU improves the model\u0026apos;s comprehension of sequential interactions in both directions, which is particularly useful in fields like foreign affairs and health that need for a more thorough contextual understanding.\u003c/p\u003e\n\u003cp\u003eTable 6: Classification Report for CNN + BiLSTM + Self-Attention\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"871\" height=\"352\"\u003e\u003c/p\u003e\n\u003cp\u003eIn Table 6 the CNN+BiLSTM+Self-Attention model performs the best, accomplished remarkable accuracy of 97% and weighted and macro averages of 0.97 for all measures. In nearly every field, this model performs admirably; in fact, certain classes (including education, science and technology, and sport) have perfect F1-scores of 1.00. With F1-scores of 0.96 and 0.91, respectively, even formerly poorer classes like Economics and Politics demonstrate significant progress. When combined with self-attention\u0026apos;s capacity to concentrate on important parts of the sequence, the BiLSTM\u0026apos;s ability in collecting long-range interdependence appears to be particularly useful for activities requiring both local and global contextual awareness. In conclusion, adding recurrent layers and attention mechanisms to CNNs leads to marked improvements in classification performance. The progression from CNN+Self-Attention to CNN+BiGRU+Self-Attention, and finally to CNN+BiLSTM+Self-Attention, reflects increasing model complexity and interpretive capacity.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo evaluate the effectiveness of the proposed attention-based hybrid deep learning models for multi-class Amharic news classification, we analyzed the confusion matrices generated for three variants: CNN with BiLSTM and self-attention, CNN with BiGRU and self-attention, and CNN with self-attention only as plotted in figure 4. These matrices provide insights into the models\u0026apos; ability to correctly classify instances across various news categories, as well as their tendency to confuse certain classes. The CNN\u003cstrong\u003e\u0026nbsp;+\u0026nbsp;\u003c/strong\u003eBiLSTM + Self-Attention model demonstrates the most robust performance among the three as indicated in\u0026nbsp;Figure 4 subfigure (a).\u0026nbsp;It achieves high accuracy across most categories, with particularly strong classification in classes such as Economy, Education, Science and Technology, and Sport, showing correct predictions of 54, 38, 52, and 41 respectively. The model also exhibits minimal misclassifications, which highlights its strength in capturing long-range dependencies and contextual information within Amharic text. However, minor confusion is observed between semantically related categories such as Politics and Foreign Affairs, which is expected given the overlapping nature of their content in real-world news reporting.\u003c/p\u003e\n\u003cp\u003eIn\u0026nbsp;Figure 4 subfigure (b),\u0026nbsp;the CNN + BiGRU + Self-Attention model also performs well, showing strong results in categories like Agriculture, Environmental, Health, and Sport. While slightly less accurate than the BiLSTM-based model, it maintains a competitive level of performance while being more computationally efficient. Nevertheless, a slightly higher degree of misclassification is noted in categories such as Politics, Culture and Tourism, and Education. This can be attributed to the lighter architecture of GRUs, which, while faster, may be less expressive in capturing complex sequential patterns compared to LSTMs. In Figure 4 subfigure (c),\u0026nbsp;illustrated\u0026nbsp;CNN + Self-Attention, which excludes the recurrent component, achieves reasonable accuracy but exhibits noticeable limitations in modeling complex class relationships. It performs well in relatively distinct categories such as Sport, Health, and Science and Technology, where the contextual boundary between classes is clearer. However, its performance drops in categories like Politics and Economy, where subtle contextual differences are crucial for accurate classification. This indicates that while self-attention enhances the model\u0026apos;s ability to focus on relevant tokens, the absence of sequential modeling (as provided by RNNs) reduces its effectiveness in capturing deep contextual meaning.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn general, the analysis confirms that the integration of CNN, bidirectional recurrent layers (BiLSTM or BiGRU), and self-attention yields significant improvements in classification performance. Among the models evaluated, the CNN + BiLSTM + Self-Attention configuration consistently provides the most accurate and reliable results, particularly in classes that require understanding of long-range dependencies.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFigure 5\u0026nbsp;presents a comparative ROC curve analysis of three different classification models evaluated on a multi-class news categorization task. Each plot illustrates the class-wise ROC curves and their corresponding AUC values across ten thematic categories: Agriculture, Culture and Tourism, Economy, Education, Environmental, Foreign Affair, Health, Politics, Science and Technology, and Sport.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(i) Left ROC Plot (\u003c/strong\u003e\u003cstrong\u003eFigure 5 sub Figure (a)\u003c/strong\u003e\u003cstrong\u003e):\u003c/strong\u003e With Culture and Tourism having the highest class-wise AUC (0.61), this plot shows comparatively balanced AUC values for the majority of categories, indicating superior separability in this region. On the other hand, Foreign Affair (AUC = 0.45) and Education (AUC = 0.36) show lower predictive ability, indicating that it may be challenging to correctly differentiate these groups from others. Limited discriminating power and possible model under-fitting are highlighted by the AUC values, which primarily range around 0.50.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;(ii) Middle ROC Plot (\u003c/strong\u003e\u003cstrong\u003eFigure 5 sub Figure (b)\u003c/strong\u003e\u003cstrong\u003e):\u003c/strong\u003e AUC values in this configuration are generally lower than in the preceding plot, with Economy and Science and Technology performing slightly at AUC = 0.53 and 0.55, respectively. With AUCs of 0.42 and 0.42, politics and education two subjects that were formerly among the weakest classes remain difficult. In many instances, the curves\u0026apos; closeness to the diagonal line indicates a performance that is almost identical to random guessing, indicating that additional optimization is required for model generalization.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(iii) Right ROC Plot (\u003c/strong\u003e\u003cstrong\u003eFigure 5 sub Figure (c)\u003c/strong\u003e\u003cstrong\u003e):\u003c/strong\u003e \u0026nbsp;Health shows the greatest AUC of 0.58, followed by Culture and Tourism (AUC = 0.56) and Environmental (AUC = 0.55) in the third ROC plot, which shows modest increases in certain classes. These indicate that class discrimination is more successful. In the meantime, the AUCs for sport and education remain lower at 0.45 and 0.44, respectively. The classifier\u0026apos;s ability to distinguish between classes is still not at its best, as indicated by the very small AUC range across every class.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn general Values mostly cluster around 0.50, indicating a consistent pattern of moderate to low AUC scores across all three models. This indicates that most categories don\u0026apos;t demonstrate high selective performance. Higher AUC scores across plots are typically obtained by classes like Culture and Tourism, Health, and Environmental, showing that these domains have more unique feature patterns. AUC values, however, are typically lower for categories like sport, foreign affairs, and education, indicating either a lack of feature representation or class overlap in the feature space. The sensitivity of the classification performance to the underlying model architecture or feature extraction method is shown by the variation in ROC curves and AUC values among the various models. In order to enhance performance in low-AUC categories, our results highlight the necessity for additional feature engineering improvement, model modification, or the integration of domain-specific knowledge.\u003c/p\u003e\n\u003cp\u003eThe accuracy of the CNN + Self-Attention model, illustrated in Figure 6 subfigure (a), increases gradually throughout the first few epochs before plateauing at around epoch 6, when it reaches a validation accuracy of roughly 0.90. The model hits an optimal point early in training, as evidenced by the validation loss flattening after epoch 4 but the training loss gradually declining. Yet, the small variation between the training and validation curves raises the possibility of minor overfitting, particularly in light of the lack of recurrent layers, which could make it more difficult for the model to accurately represent temporal dependencies in the input. The CNN + BiGRU + Self-Attention model, on the other hand, exhibits significantly faster convergence, with both training and validation accuracy above 0.90 during the first three epochs, as shown in Figure 6 subfigure (b). Indicating that the model learns rapidly and generalizes effectively, the training and validation losses decrease dramatically at the beginning and level off at epoch 4. The reduced difference between the accuracy and loss training and validation curves indicates that the BiGRU part improves the ability of the model to handle temporal information while maintaining the model\u0026apos;s relative light weight and efficiency.\u003c/p\u003e\n\u003cp\u003eA comparable trend can be noticed in the CNN + BiLSTM + Self-Attention model, which is shown in Figure 6 subfigure (c). It achieves high accuracy values above 0.95 in a short period of time and has the lowest training and validation losses of the three models. High generalization is showed by the fact that its validation performance constantly approaches training performance. The focus mechanism of self-attention and BiLSTM\u0026apos;s capability to identify long-range dependencies in both directions work well together in this paradigm. It provides the best convergence speed and stability, despite being somewhat more computationally complex than BiGRU.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo assess the interpretability and decision rationale of the classification model, LIME (Local Interpretable Model-Agnostic Explanations) was employed to visualize the most influential words contributing to predictions within the Health category.\u003c/p\u003e\n\u003cp\u003eFigure 7 presents three representative examples where the model correctly classified input texts as health-related with high confidence.The model assigned a prediction probability of 0.88 for the Health class, with minor probabilities assigned to Science and Technology (0.04), Economy (0.02), and Environmental (0.02). The highlighted words such as \u0026ldquo;ሕክምና\u0026rdquo; (treatment), \u0026ldquo;እንክብካቤ\u0026rdquo; (care), and \u0026ldquo;የጤና\u0026rdquo; (of health) illustrate the model\u0026rsquo;s focus on semantically relevant tokens, reinforcing the correctness of the prediction.\u003c/p\u003e\n\u003cp\u003eIn Figure 8 the model\u0026apos;s prediction for Health increased to 0.98, with negligible competition from other classes. The explanation highlights terms like \u0026ldquo;በሽታ\u0026rdquo; (disease), \u0026ldquo;ክስተት\u0026rdquo; (crisis), and \u0026ldquo;አካል\u0026rdquo; (body), all of which are core to the health domain. This near-perfect confidence suggests the model has developed a strong internal representation of medical and wellness-related language.\u003c/p\u003e\n\u003cp\u003eEven with a slightly lower confidence level (0.93), Figure 9 indicates that Health still has a significant advantage over rival classes like Politics (0.05) and Foreign Affairs (0.01). Consistent health-related phrases like \u0026ldquo;የጤና\u0026rdquo;, \u0026ldquo;ሕክምና\u0026rdquo;, and \u0026ldquo;እንክብካቤ\u0026rdquo;, are highlighted by the model once more, indicating thematic consistency and robustness across samples. The model\u0026apos;s predictions are based on significant linguistic features rather than meaningless word patterns, as the LIME visualizations generally reflect. The model has successfully internalized crucial semantic structures related to the health category, as seen by the periodic presence of domain-specific terms.\u003c/p\u003e"},{"header":"5. Conclusions","content":"\u003cp\u003eA significant achievement in the field of explainable AI (XAI) and natural language processing (NLP) of underrepresented languages is the building and use of an attention-based hybrid deep learning model for multi-class Amharic news classification. Utilizing the blending of various neural network components and attention interprets that improve performance and interpretability, this model successfully tackles the particular linguistic challenges of the Amharic language, including complicated morphology, agglutinative structure, and sparse annotated data. A hybrid neural network, which usually integrates Convolutional Neural Networks (CNNs) along with Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Units (BiGRU), occupies the core of the architecture. Whereas the BiLSTM/BiGRU layers are adept at modeling long-range relationships in text sequences, the CNN layers are outstanding at extracting local and position-invariant information. The model's ability to selectively focus on semantically relevant words and phrases is further enhanced by the integration of a self-attention mechanism. Self-attention enables the model to offer more reliable and contextually aware representations of the input by giving relevant tokens more weight, regardless of where they are in the text. Notably, the system's transparency and reliability are improved by integrating Explainable AI (XAI) tools, particularly LIME (Local Interpretable Model-agnostic Explanations). This becomes especially essential in critical or high-stakes fields like the classification of legal or political news, where user trust and ethical AI adoption depend on explainability. In short, this attention-based hybrid model is scalable and interpretable, which makes it suitable for other low-resource languages and domains in addition to achieving state-of-the-art performance in multi-class Amharic news classification. It opens the door for more accountable, intelligible, and efficient NLP solutions for Amharic and beyond by integrating explainability frameworks with powerful deep learning architectures.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e Authors declare no funding for this research\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e The authors declare that there is no competing interest\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of interest:\u003c/strong\u003e The authors declare that there is no conflict of interest\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial:\u003c/strong\u003e not applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics, Consent to Participate, and Consent to Publish declarations:\u003c/strong\u003e not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data:\u003c/strong\u003e The data generated during and/or analysed the current study are not publicly available but are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contribution:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAmlakie Aschale Alemu:\u003c/strong\u003e\u0026nbsp; Data curation, Visualization, Validation, Methodology, Writing, Reviewing and Editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMalefia Demilie Melesse:\u003c/strong\u003e Visualization, Methodology, Reviewing and Editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDaniel Arega Mengesha\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e Conceptualization, Visualization, and document correction.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMisganaw Aguate Widneh:\u003c/strong\u003e Conceptualization, Formal analysis, Investigation\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eD. E. Id, G. Haile, and W. Taye, \u0026ldquo;recurrent unit model for Amharic next word prediction,\u0026rdquo; pp. 1\u0026ndash;10, 2022, doi: 10.1371/journal.pone.0273156.\u003c/li\u003e\n\u003cli\u003eB. M. Hailu, \u0026ldquo;Semantic Role Labeling for Amharic Text Using Multiple Embeddings and Deep Neural Network,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 11, no. March, pp. 33274\u0026ndash;33295, 2023, doi: 10.1109/ACCESS.2023.3263147.\u003c/li\u003e\n\u003cli\u003eN. A. Asfaw, B. H. Belay, and K. M. Alemu, \u0026ldquo;OPEN A deep learning approach for line-level Amharic Braille image recognition,\u0026rdquo; pp. 1\u0026ndash;14, 2024.\u003c/li\u003e\n\u003cli\u003eD. Endalie and G. Haile, \u0026ldquo;Automated Amharic News Categorization Using Deep,\u0026rdquo; vol. 2021, 2021, doi: 10.1155/2021/3774607.\u003c/li\u003e\n\u003cli\u003eC. Science, C. Science, C. Science, and C. Science, \u0026ldquo;Explainable Machine Learning For Text Classification : A Novel Approach To Transparency And Interpretability,\u0026rdquo; vol. 103, no. 3, pp. 907\u0026ndash;915, 2025.\u003c/li\u003e\n\u003cli\u003eS. Gao, B. He, and F. Schilder, \u0026ldquo;On the Effectiveness of Pre-Trained Language Models for Legal Natural Language Processing : An Empirical Study,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 10, no. May, pp. 75835\u0026ndash;75858, 2022, doi: 10.1109/ACCESS.2022.3190408.\u003c/li\u003e\n\u003cli\u003eA. A. Fenta and S. Gebeyehu, \u0026ldquo;Automatic Idiom Identification Model for Amharic,\u0026rdquo; vol. 22, no. 8, 2023, doi: 10.1145/3606864.\u003c/li\u003e\n\u003cli\u003eM. Degu, A. Olalekan, G. Kassa, and T. Bezabh, \u0026ldquo;Contextual word disambiguates of Ge \u0026rsquo; ez language with homophonic using machine learning,\u0026rdquo; \u003cem\u003eAmpersand\u003c/em\u003e, vol. 12, no. December 2023, p. 100169, 2024, doi: 10.1016/j.amper.2024.100169.\u003c/li\u003e\n\u003cli\u003eB. H. Belay, G. B. Gebremeskel, B. B. Bezabih, and S. Gebeyehu, \u0026ldquo;Factorized Recurrent Neural Network with Attention for,\u0026rdquo; vol. 22, no. 12, 2023, doi: 10.1145/3630607.\u003c/li\u003e\n\u003cli\u003eE. F. Dinsa, M. Das, T. Urgessa, and A. Krishnaraj, \u0026ldquo;Automatic categorization of medical documents in Afaan Oromo using ensemble machine learning techniques,\u0026rdquo; \u003cem\u003eDiscov. Appl. Sci.\u003c/em\u003e, 2024, doi: 10.1007/s42452-024-06307-0.\u003c/li\u003e\n\u003cli\u003eK. Wang, Y. Ding, and S. C. Han, \u003cem\u003eGraph neural networks for text classification : a survey\u003c/em\u003e, vol. 57, no. 8. Springer Netherlands, 2024. doi: 10.1007/s10462-024-10808-0.\u003c/li\u003e\n\u003cli\u003eD. Voskergian, R. Jayousi, and M. Yousef, \u0026ldquo;Topic selection for text classification using ensemble topic modeling with grouping , scoring , and modeling approach,\u0026rdquo; pp. 1\u0026ndash;19, 2024.\u003c/li\u003e\n\u003cli\u003eT. Thaminkaew, P. Lertvittayakumjorn, and P. Vateekul, \u0026ldquo;Prompt-Based Label-Aware Framework for Few-Shot Multi-Label Text Classification,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 12, no. January, pp. 28310\u0026ndash;28322, 2024, doi: 10.1109/ACCESS.2024.3367994.\u003c/li\u003e\n\u003cli\u003eK. Taha, P. D. Yoo, C. Yeun, D. Homouz, and A. Taha, \u0026ldquo;A comprehensive survey of text classification techniques and their research applications : Observational and experimental insights,\u0026rdquo; \u003cem\u003eComput. Sci. Rev.\u003c/em\u003e, vol. 54, no. July, p. 100664, 2024, doi: 10.1016/j.cosrev.2024.100664.\u003c/li\u003e\n\u003cli\u003eH. Sakai and S. S. Lam, \u0026ldquo;Large Language Models For Health Care Text Classification : A Systematic Review arXiv : 2503 . 01159v1 [ cs . CL ] 3 Mar 2025\u0026rdquo;.\u003c/li\u003e\n\u003cli\u003eT. Kuzman, \u0026ldquo;LLM Teacher-Student Framework for Text Classification With No Manually Annotated Data : A Case Study in IPTC News Topic Classification,\u0026rdquo; vol. 13, no. January, pp. 35621\u0026ndash;35633, 2025, doi: 10.1109/ACCESS.2025.3544814.\u003c/li\u003e\n\u003cli\u003eA. Jahin, S. H. Shovon, M. F. Mridha, and R. Islam, \u0026ldquo;OPEN A hybrid transformer and attention based recurrent neural network for robust and interpretable sentiment analysis of tweets,\u0026rdquo; pp. 1\u0026ndash;25, 2024.\u003c/li\u003e\n\u003cli\u003eS. Hirpassa and G. S. Lehal, \u0026ldquo;Heliyon Improving part-of-speech tagging in Amharic language using deep neural network,\u0026rdquo; \u003cem\u003eHeliyon\u003c/em\u003e, vol. 9, no. 7, p. e17175, 2023, doi: 10.1016/j.heliyon.2023.e17175.\u003c/li\u003e\n\u003cli\u003eS. Gite, H. Khatavkar, K. Kotecha, and S. Srivastava, \u0026ldquo;Explainable stock prices prediction from financial news articles using sentiment analysis,\u0026rdquo; pp. 1\u0026ndash;21, 2021, doi: 10.7717/peerj-cs.340.\u003c/li\u003e\n\u003cli\u003eA. Nasayreh \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Journal of King Saud University - Computer and Jordanian banknote data recognition : A CNN-based approach with attention mechanism,\u0026rdquo; \u003cem\u003eJ. King Saud Univ. - Comput. Inf. Sci.\u003c/em\u003e, vol. 36, no. 4, p. 102038, 2024, doi: 10.1016/j.jksuci.2024.102038.\u003c/li\u003e\n\u003cli\u003eS. Khan \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;BiCHAT : BiLSTM with deep CNN and hierarchical attention for hate speech detection,\u0026rdquo; \u003cem\u003eJ. King Saud Univ. - Comput. Inf. Sci.\u003c/em\u003e, vol. 34, no. 7, pp. 4335\u0026ndash;4344, 2022, doi: 10.1016/j.jksuci.2022.05.006.\u003c/li\u003e\n\u003cli\u003eZ. Dai and Y. He, \u0026ldquo;A non-functional requirements classification model based on cooperative attention mechanism fused with label embedding,\u0026rdquo; \u003cem\u003eComput. Electr. Eng.\u003c/em\u003e, vol. 121, no. September 2024, p. 109856, 2025, doi: 10.1016/j.compeleceng.2024.109856.\u003c/li\u003e\n\u003cli\u003eC. Rudin, \u0026ldquo;Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,\u0026rdquo; \u003cem\u003eNat. Mach. Intell.\u003c/em\u003e, vol. 1, no. 5, pp. 206\u0026ndash;215, 2019, doi: 10.1038/s42256-019-0048-x.\u003c/li\u003e\n\u003cli\u003eJ. R. Jim, M. A. R. Talukder, P. Malakar, M. M. Kabir, K. Nur, and M. F. Mridha, \u0026ldquo;Recent advancements and challenges of NLP-based sentiment analysis: A state-of-the-art review,\u0026rdquo; \u003cem\u003eNat. Lang. Process. J.\u003c/em\u003e, vol. 6, no. January, p. 100059, 2024, doi: 10.1016/j.nlp.2024.100059.\u003c/li\u003e\n\u003cli\u003eV. Buhrmester, D. M\u0026uuml;nch, and M. Arens, \u0026ldquo;Analysis of Explainers of Black Box Deep Neural Networks for Computer Vision: A Survey,\u0026rdquo; \u003cem\u003eMach. Learn. Knowl. Extr.\u003c/em\u003e, vol. 3, no. 4, pp. 966\u0026ndash;989, 2021, doi: 10.3390/make3040048.\u003c/li\u003e\n\u003cli\u003eQ. Xu \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Applications of Explainable AI in Natural Language Processing,\u0026rdquo; \u003cem\u003eGlob. Acad. Front.\u003c/em\u003e, vol. 2, no. 3, pp. 51\u0026ndash;64, 2024.\u003c/li\u003e\n\u003cli\u003eS. Luo, H. Ivison, S. C. Han, and J. Poon, \u0026ldquo;Local Interpretations for Explainable Natural Language Processing: A Survey,\u0026rdquo; \u003cem\u003eACM Comput. Surv.\u003c/em\u003e, vol. 56, no. 9, 2024, doi: 10.1145/3649450.\u003c/li\u003e\n\u003cli\u003eE. Hashmi, S. Y. Yayilgan, M. M. Yamin, S. Ali, and M. Abomhara, \u0026ldquo;Advancing Fake News Detection : Hybrid Deep Learning With FastText and Explainable AI,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 12, no. March, pp. 44462\u0026ndash;44480, 2024, doi: 10.1109/ACCESS.2024.3381038.\u003c/li\u003e\n\u003cli\u003eS. M. Carta, S. Consoli, L. Piras, A. S. Podda, and D. R. Recupero, \u0026ldquo;Explainable Machine Learning Exploiting News and Domain-Specific Lexicon for Stock Market Forecasting,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 9, pp. 30193\u0026ndash;30205, 2021, doi: 10.1109/ACCESS.2021.3059960.\u003c/li\u003e\n\u003cli\u003eM. M. Azad and H. S. Kim, \u0026ldquo;An explainable artificial intelligence-based approach for reliable damage detection in polymer composite structures using deep learning,\u0026rdquo; \u003cem\u003ePolym. Compos.\u003c/em\u003e, no. September 2024, pp. 1536\u0026ndash;1551, 2024, doi: 10.1002/pc.29055.\u003c/li\u003e\n\u003cli\u003eI. Of, D. Learning, M. In, and N. Language, \u0026ldquo;Article Info,\u0026rdquo; vol. 3, no. 2, pp. 56\u0026ndash;66, 2025.\u003c/li\u003e\n\u003cli\u003eM. S. H. Apu, M. S. Islam, and T. T. Aurpa, \u0026ldquo;Explainable AI for Sentiment Analysis of Human Metapneumovirus (HMPV) Using XLNet,\u0026rdquo; 2025, [Online]. Available: http://arxiv.org/abs/2502.01663\u003c/li\u003e\n\u003cli\u003eN. Pospelova, A. Tarasova, N. Subbotina, N. Koroleva, N. Raimova, and E. L. Lydia, \u0026ldquo;Explainable Artificial Intelligence and Natural Language Processing for Unraveling Deceptive Contents,\u0026rdquo; \u003cem\u003eFusion Pract. Appl.\u003c/em\u003e, vol. 14, no. 2, pp. 146\u0026ndash;158, 2024, doi: 10.54216/FPA.140212.\u003c/li\u003e\n\u003cli\u003eK. Zahoor, N. Z. Bawany, and T. Qamar, \u0026ldquo;Evaluating text classification with explainable artificial intelligence,\u0026rdquo; \u003cem\u003eIAES Int. J. Artif. Intell.\u003c/em\u003e, vol. 13, no. 1, pp. 278\u0026ndash;286, 2024, doi: 10.11591/ijai.v13.i1.pp278-286.\u003c/li\u003e\n\u003cli\u003eT. Bao, N. Ren, R. Luo, B. Wang, G. Shen, and T. Guo, \u0026ldquo;A BERT-Based Hybrid Short Text Classification Model Incorporating CNN and Attention-Based BiGRU,\u0026rdquo; \u003cem\u003eJ. Organ. End User Comput.\u003c/em\u003e, vol. 33, no. 6, pp. 1\u0026ndash;21, 2021, doi: 10.4018/joeuc.294580.\u003c/li\u003e\n\u003cli\u003eM. Ahmed, M. S. Hossain, R. Ul Islam, and K. Andersson, \u0026ldquo;Explainable Text Classification Model for COVID-19 Fake News Detection,\u0026rdquo; \u003cem\u003eJ. Internet Serv. Inf. Secur.\u003c/em\u003e, vol. 12, no. 2, pp. 51\u0026ndash;69, 2022, doi: 10.22667/JISIS.2022.05.31.051.\u003c/li\u003e\n\u003cli\u003eA. Caleb, \u0026ldquo;Multi-Class Fake News Detection Using Deep Learning and Explainable AI : A Transformer-Gradient Boosting Approach,\u0026rdquo; no. March, 2025.\u003c/li\u003e\n\u003cli\u003eM. Cesarini, L. Malandri, F. Pallucchini, A. Seveso, and F. Xing, \u0026ldquo;Explainable AI for Text Classification: Lessons from a Comprehensive Evaluation of Post Hoc Methods,\u0026rdquo; \u003cem\u003eCognit. Comput.\u003c/em\u003e, pp. 3077\u0026ndash;3095, 2024, doi: 10.1007/s12559-024-10325-w.\u003c/li\u003e\n\u003cli\u003eB. Jang, M. Kim, G. Harerimana, S. Kang, and J. W. Kim, \u0026ldquo;applied sciences Bi-LSTM Model to Increase Accuracy in Text Classification : Combining Word2vec CNN and Attention Mechanism,\u0026rdquo; 2020.\u003c/li\u003e\n\u003cli\u003eJ. Blake, A. B. U. Saleh, M. Miah, K. Kredens, J. Shin, and S. Member, \u0026ldquo;Detection of AI-Generated Texts : A Bi-LSTM and Attention-Based Approach,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 13, no. April, pp. 71563\u0026ndash;71576, 2025, doi: 10.1109/ACCESS.2025.3562750.\u003c/li\u003e\n\u003cli\u003eM. Yibeyin, Y. Tehone, A. Liyih, and M. Fentie, \u0026ldquo;Public opinion mining in social media about Ethiopian broadcasts using deep learning,\u0026rdquo; pp. 1\u0026ndash;10, 2024.\u003c/li\u003e\n\u003cli\u003eM. M. Hasan \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Hybrid deep learning: a comparative study on ai algorithms in natural language processing for text classification,\u0026rdquo; \u003cem\u003eBull. Electr. Eng. Informatics\u003c/em\u003e, vol. 14, no. 1, pp. 551\u0026ndash;559, 2025, doi: 10.11591/eei.v14i1.7617.\u003c/li\u003e\n\u003cli\u003eS. Murindanyi, B. A. Yiiki, A. Katumba, and J. Nakatumba-Nabende, \u0026ldquo;Explainable Machine Learning Models for Swahili News Classification,\u0026rdquo; \u003cem\u003eACM Int. Conf. Proceeding Ser.\u003c/em\u003e, pp. 12\u0026ndash;18, 2023, doi: 10.1145/3639233.3639250.\u003c/li\u003e\n\u003cli\u003eD. Rohera \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;A Taxonomy of Fake News Classification Techniques: Survey and Implementation Aspects,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 10, pp. 30367\u0026ndash;30394, 2022, doi: 10.1109/ACCESS.2022.3159651.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-applied-sciences","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Applied Sciences](https://link.springer.com/journal/42452)","snPcode":"42452","submissionUrl":"https://submission.springernature.com/new-submission/42452/3","title":"Discover Applied Sciences","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Explainable AI, Self_Attention, CNN + BiGRU, CNN + BiLSTM, LIME","lastPublishedDoi":"10.21203/rs.3.rs-6759877/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6759877/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eEfficient and adaptable text classification systems are required due to the increasing expansion of Amharic electronic documents in order to manage and extract important insights from large amounts of data. Traditional natural language processing (NLP) methods are severely hampered by the Amharic language's intricate morphology and sparse annotated corpora. By combining the advantages of Bidirectional GRU and Bidirectional LSTM networks for capturing sequential dependencies and Convolutional Neural Networks (CNN) for local feature extraction, this study proposes a novel attention-based hybrid deep learning model for multi-class Amharic news classification. The model's discriminative power across several categories is improved by incorporating a self-attention mechanism that dynamically emphasizes contextually significant terms. We employ Explainable AI (XAI) methods, like Local Interpretable Model-Agnostic Explanations (LIME), which offer human-interpretable explanations for predictions, to minimize the hidden nature of deep learning decisions and enhance trust in AI systems. A carefully selected multi-class Amharic news dataset that encompasses a variety of topics, including agriculture, culture and tourism, education, the economy, the environment, foreign affairs, health, politics, science and technology, and sports, is used to evaluate the model. The experimental results show that CNN\u0026thinsp;+\u0026thinsp;BiLSTM with Self-Attention scores 96.7, 96.8, 96.8 and 97 for precision, recall, F1-score, and accuracy respectively. This study introduces to the larger objective of transparent and reliable AI in low-resource language contexts and pushes beyond the limits of Amharic automatic language processing.\u003c/p\u003e","manuscriptTitle":"Attention Based Hybrid Deep Learning models for Multi class Amharic News Categorization with Explainable AI","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-09 15:42:53","doi":"10.21203/rs.3.rs-6759877/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-24T18:06:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-19T10:42:31+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"120698434480608130072551093797633179757","date":"2025-06-19T10:02:46+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-16T17:11:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"132257185480058340854788560186978355677","date":"2025-06-10T13:28:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"238016454633511636777279299027731559374","date":"2025-06-10T12:58:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"306351606877405263381867644829167398383","date":"2025-06-10T09:57:41+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-06-06T08:25:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-03T09:03:23+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-03T09:03:15+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Applied Sciences","date":"2025-05-27T13:18:59+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-applied-sciences","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Applied Sciences](https://link.springer.com/journal/42452)","snPcode":"42452","submissionUrl":"https://submission.springernature.com/new-submission/42452/3","title":"Discover Applied Sciences","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5d3a1007-a3ea-4593-ae54-a4e36e022ac1","owner":[],"postedDate":"June 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-12-10T07:39:11+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-09 15:42:53","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6759877","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6759877","identity":"rs-6759877","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}