Deep and Contextually Engineered Features for Metaphor Detection

Abstract The focal point of this work is to automatically detect metaphor instances in short texts. It is the study of extricating the most optimal features for the task by using a deep learning architecture combined with carefully handcrafted contextual features. All of these will be discussed in detail in this paper. It is also found that a few sets performed well when they are used independently, but the others not so much. However, even the latter sets become very useful after the combination process with the former sets. Finally, the combined feature sets undergo the classification using well-known machine learning classification algorithms. It is decided that all the five algorithms are used for the purpose of comparison. The best algorithm for this task is found to be Support Vector Machine (SVM). The outcome of all the experiments are good in all the metrics used. Furthermore, result comparison in terms of F1-measure to existing works in the same domain is compiled and stated in this paper. The comparison on each feature set is also stated as additional information.