AB-TC-BLATT: A Resource-Efficient Parallel System Architecture with Frozen ALBERT for Practical Chinese Sentiment Analysis

AB-TC-BLATT: A Resource-Efficient Parallel System Architecture with Frozen ALBERT for Practical Chinese Sentiment Analysis | 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 AB-TC-BLATT: A Resource-Efficient Parallel System Architecture with Frozen ALBERT for Practical Chinese Sentiment Analysis Li Qiusheng, Long Yu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8876236/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The exponential growth of user-generated reviews on e-commerce and social media platforms has made sentiment analysis an indispensable tool for mining consumer insights. However, performing accurate sentiment analysis on Chinese text presents unique challenges, including word segmentation ambiguity and pervasive polysemy, which are often exacerbated in practical deployment scenarios characterized by limited labeled data or constrained computational budgets. While pre-trained models like ALBERT provide powerful contextual representations, their conventional serial integration with downstream neural networks for sentiment analysis not only risks creating an information bottleneck but also leads to high computational cost and potential overfitting when training data is scarce. To address these practical issues for system deployment, this paper proposes AB-TC-BLATT, a resource-efficient and practically-oriented model featuring a novel dual-channel parallel architecture. Our core design is to keep the ALBERT parameters frozen to preserve general linguistic knowledge and ensure training efficiency, while a parallel, lightweight local channel employs trainable embeddings with TextCNN and BiLSTM-Attention to adaptively capture task-specific and Chinese-specific linguistic patterns from the available, often limited, data. A hierarchical fusion strategy then integrates these complementary features. Comprehensive experiments and a thorough system efficiency analysis on two Chinese review datasets demonstrate that AB-TC-BLATT achieves superior accuracy (94.38% and 91.43%). Crucially, it maintains robust performance under simulated low-resource conditions and exhibits strong generalization potential. More importantly, from a system engineering perspective, it significantly reduces trainable parameters (by 63.5%) and training time (by 52.8%) compared to serial fusion counterparts. The model offers an effective, efficient, and practical paradigm for building scalable and low-resource sentiment analysis systems, providing a reusable architectural template particularly suited for resource-constrained Chinese language applications. Sentiment Analysis ALBERT Resource-Efficient NLP System Efficiency Parallel Neural Network Architecture Hierarchical Feature Fusion Chinese Text Classification Low-Resource Text Classification Frozen Pre-trained Models Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted 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-8876236","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":591585077,"identity":"a8ae71f8-39e7-4bce-91cf-e66b143cfe86","order_by":0,"name":"Li Qiusheng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8ElEQVRIie3RrwoCMRzA8Z8MZtlcPYu+wkRQxOCrnBgshvMJNM2iXYNvYDgRxHjyqyfWCwZBuKKCYBIuqOe/tjMK7stgY+zDBgMwmX6wDAHw4gHEe+7ZekI/hL6OJpHHdCdMfknSZIos2uTKYnLetxcIIt2ScFnoHkYd5CosVobHeXXkI2T7B5ka+DrCJPIu1t1gNS9yhSCDliQplUBYhB038MOY1L4jFG257pNdfIuVSKizHCssuAEtEa6azPJDZznQECFwdjpEmJdr3J25quZErzHdXjTk1vM7LJtat4nd154WvInwyCnhpMlkMv1pV6I+UrshW7oPAAAAAElFTkSuQmCC","orcid":"","institution":"Gannan Normal University","correspondingAuthor":true,"prefix":"","firstName":"Li","middleName":"","lastName":"Qiusheng","suffix":""},{"id":591585078,"identity":"8fb4d667-4f78-447a-92d4-d0726c4fa744","order_by":1,"name":"Long Yu","email":"","orcid":"","institution":"Gannan Normal University","correspondingAuthor":false,"prefix":"","firstName":"Long","middleName":"","lastName":"Yu","suffix":""}],"badges":[],"createdAt":"2026-02-14 02:08:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8876236/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8876236/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106404540,"identity":"ef2d3e9b-2c41-47bd-82b9-ab28f1237322","added_by":"auto","created_at":"2026-04-08 09:16:12","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":556905,"visible":true,"origin":"","legend":"","description":"","filename":"ABTCBLATT.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8876236/v1_covered_9cb2199e-b783-4503-9631-2960e6a5c234.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"AB-TC-BLATT: A Resource-Efficient Parallel System Architecture with Frozen ALBERT for Practical Chinese Sentiment Analysis","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Sentiment Analysis, ALBERT, Resource-Efficient NLP, System Efficiency, Parallel Neural Network Architecture, Hierarchical Feature Fusion, Chinese Text Classification, Low-Resource Text Classification, Frozen Pre-trained Models","lastPublishedDoi":"10.21203/rs.3.rs-8876236/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8876236/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe exponential growth of user-generated reviews on e-commerce and social media platforms has made sentiment analysis an indispensable tool for mining consumer insights. However, performing accurate sentiment analysis on Chinese text presents unique challenges, including word segmentation ambiguity and pervasive polysemy, which are often exacerbated in practical deployment scenarios characterized by limited labeled data or constrained computational budgets. While pre-trained models like ALBERT provide powerful contextual representations, their conventional serial integration with downstream neural networks for sentiment analysis not only risks creating an information bottleneck but also leads to high computational cost and potential overfitting when training data is scarce. To address these practical issues for system deployment, this paper proposes AB-TC-BLATT, a resource-efficient and practically-oriented model featuring a novel dual-channel parallel architecture. Our core design is to keep the ALBERT parameters frozen to preserve general linguistic knowledge and ensure training efficiency, while a parallel, lightweight local channel employs trainable embeddings with TextCNN and BiLSTM-Attention to adaptively capture task-specific and Chinese-specific linguistic patterns from the available, often limited, data. A hierarchical fusion strategy then integrates these complementary features. Comprehensive experiments and a thorough system efficiency analysis on two Chinese review datasets demonstrate that AB-TC-BLATT achieves superior accuracy (94.38% and 91.43%). Crucially, it maintains robust performance under simulated low-resource conditions and exhibits strong generalization potential. More importantly, from a system engineering perspective, it significantly reduces trainable parameters (by 63.5%) and training time (by 52.8%) compared to serial fusion counterparts. The model offers an effective, efficient, and practical paradigm for building scalable and low-resource sentiment analysis systems, providing a reusable architectural template particularly suited for resource-constrained Chinese language applications.\u003c/p\u003e","manuscriptTitle":"AB-TC-BLATT: A Resource-Efficient Parallel System Architecture with Frozen ALBERT for Practical Chinese Sentiment Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-08 05:35:46","doi":"10.21203/rs.3.rs-8876236/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9283ca3d-025e-437c-9902-310c8ff52a91","owner":[],"postedDate":"April 8th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-08T05:35:46+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-08 05:35:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8876236","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8876236","identity":"rs-8876236","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}