Citation :

Sumy T O, Vinoth A, "Deep Contextual Feature Fusion with CFF-Net for Tamil and Malayalam Sentiment Analysis," International Journal of Engineering Trends and Technology (IJETT), vol. 74, no. 9, pp. 113-127, 2026. Crossref, https://doi.org/10.14445/22315381/IJETT-V74I9P110

Abstract

Sentiment analysis of the Dravidian languages, such as Tamil and Malayalam, is necessary in most cases to decipher the perception and sentiments expressed in social media, which is normally rich in code-mixed, informal, and vague text. The issue of noisy, skewed and context-variable data in the context of accurate classification is a problem in these languages. The proposed solution to these challenges is a new sentiment classification model that incorporates Contextual Feature Fusion Network (CFF-Net) with Adaptive Gradient-Weighted Optimization (AGWO) to overcome these challenges. The aim is to enhance accuracy in classifications, address the issue of class imbalance and provide strong detection of positive, negative, and neutral sentiments. The framework identifies context-sensitive features of the text in the form of hierarchical levels of attention and semantic interaction, and focuses on informative tokens, but rejects irrelevant information. AGWO automatically selects the learning rates of each modality, which converges faster and more steadily. Tests were run on complete datasets of Tamil and Malayalam, and performance was measured in terms of Accuracy, Balanced Accuracy, Precision, Recall, F1-Score, and AUC. The findings indicate that the proposed CFF-Net with AGWO is superior to the current models in all metrics, with an accuracy of 97.14% on Tamil and 96.67% on Malayalam, and one can conclude that it is effective in sentiment classification and code-mixed and imbalanced text. On the whole, this framework is able to view as an effective and reliable tool in the sentiment analysis of the Dravidian languages, and it is used as the basis of future studies in the field of multilingual text comprehension and in-time social media analytics.

Keywords

Adaptive Gradient-Weighted Optimization, Contextual Feature Fusion Network, Code-Mixed text, Dravidian languages, Sentiment classification.

References

[1] Muhammad Abbas et al., Multinomial Naive Bayes Classification Model for Sentiment Analysis,” International Journal of Computer Science and Network Security, vol. 19, no. 3, pp. 62-70, 2019.
[Google Scholar]

[2] Abdullah Al Nahian et al., “NLPopsCIOL@DravidianLangTech 2025: Classification of Abusive Tamil and Malayalam text Targeting Women using Pre-Trained Models,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 38-45, 2025. 
[CrossRef] [Google Scholar] [Publisher Link]

[3] P. Ambily Pramitha, and John T Abraham, “Hybrid Classifier for Sentiment Analysis in Malayalam with Modified TF-IDF Features,” International Journal of Modeling, Simulation, and Scientific Computing, vol. 14, no. 5, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[4] Anitha R et al., “Enhancing Trust and Interpretability in Malayalam Sentiment Analysis with Explainable AI,” Proceedings of the 21^st International Conference on Natural Language Processing (ICON), NLP Association of India (NLPAI), pp 102-108, 2024.
[Google Scholar] [Publisher Link]

[5] M. Arunmozhi et al., “MADTRAS: Dataset for Aspect-based Sentiment Analysis of Movie Reviews in Tamil,” Data in Brief, vol. 63, pp. 1-7, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[6] Ponsubash Raj R, Paruvatha Priya B, and Bharathi B, “JustATalentedTeam@DravidianLangTech 2025: A Study of ML and DL Approaches for Sentiment Analysis in Code-Mixed Tamil and Tulu Texts,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 273-277, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[7] Bharathi Raja Chakravarthi, “Sarcasm Detection in Tamil and Malayalam YouTube Comments,” Social Network Analysis and Mining, vol. 15, no. 1, pp. 1-18, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[8] Bharathi Raja Chakravarthi et al., “A Sentiment Analysis Dataset for Code-Mixed Malayalam-English,” Proceedings of the 1^st Joint Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU) and Collaboration and Computing for Under-Resourced Languages (CCURL), European Language Resources Association, pp 177-184, 2020. [Google Scholar] [Publisher Link]

[9] Supriya Chanda, Anshika Mishra, and Sukomal Pal, “Sentiment analysis of Code-Mixed Dravidian Languages Leveraging Pretrained Model and Word-Level Language Tag,” Natural Language Processing, vol. 31, no. 2, pp. 477-499, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[10] Kannaiah Chattun et al., “Sentiment Classification for Telugu using Transformed based Approaches on a Multi-Domain Dataset,” Scientific Reports, vol. 15, no. 1, pp. 1-21, 2025.
[CrossRef] [Google Scholar] [Publisher Link] 

[11] Md. Sajid Alam Chowdhury et al., “Fired_from_NLP@DravidianLangTech 2025: A Multimodal Approach for Detecting Misogynistic Content in Tamil and Malayalam Memes,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 459-464, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[12] Syam Mohan Elankath, and Sunitha Ramamirtham, “Sentiment Analysis of Malayalam Tweets using Bidirectional Encoder Representations from Transformers: A Study,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 29, no. 3, pp. 1817-1826, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[13] Enjamamul Haque Eram et al., “Eureka-CIOL@DravidianLangTech 2025: Using Customized BERTs for Sentiment Analysis of Tamil Political Comments,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 6-11, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[14] Nikhil Ghanghor et al., “IIITK@DravidianLangTech-EACL2021: Offensive Language Identification and meme Classification in Tamil, Malayalam, and Kannada,” Proceedings of the First Workshop on Speech and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 222-229, 2021.
[Google Scholar] [Publisher Link]

[15] Zabit Hameed, and Begonya Garcia-Zapirain, “Sentiment Classification using a Single-Layered BiLSTM Model,” IEEE Access vol. 8, pp. 73992-74001, 2020.
[CrossRef] [Google Scholar] [Publisher Link]

[16] Jisha P. Jayan, J. Satheesh Kumar, and T. Amudha, “Semantic role Identification for Malayalam using Machine Learning Approaches,” Innovations in Systems and Software Engineering, vol. 21, no. 1, pp. 279-285, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[17] Abirami Jayaraman et al., “AnalysisArchitects@DravidianLangTech 2025: Machine Learning Approach to Political Multiclass Sentiment Analysis of Tamil,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 614-618, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[18] Sandra Johnson, Boomika E, and Lahari P, “RMKMavericks@DravidianLangTech 2025: Tackling Abusive Tamil and Malayalam text Targeting Women: A Linguistic Approach,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp. 19-23, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[19] Adaikkan Kalaivani, and Durairaj Thenmozhi, “Multilingual Sentiment Analysis in Tamil, Malayalam, and Kannada Code-Mixed Social Media Posts using MBERT,” FIRE, pp. 1020-1028, 2021.
[Google Scholar]

[20] Shreyas Karthik et al., “Sarcasm Identification of Dravidian Languages Malayalam and Tamil,” FIRE, pp. 310-317, 2024.
[Google Scholar]

[21] Abhinav Kumar, Sunil Saumya, and Jyoti Prakash Singh, “An Ensemble-based Model for Sentiment Analysis of Dravidian Code-Mixed Social Media Posts,” FIRE, pp 950-958, 2021.
[Google Scholar]

[22] Niranjan Kumar et al., “LexiLogic@DravidianLangTech 2025: Detecting Misogynistic Memes and Abusive Tamil and Malayalam text Targeting Women on Social Media,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 435-439, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[23] Durga Prasad Manukonda, Rohith Gowtham Kodali, and Daniel Iglesias, byteSizedLLM@DravidianLangTech 2025: Multimodal Hate Speech Detection in Malayalam using Attention-Driven BiLSTM, Topic-BERT, and Wav2Vec 2.0, Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 68-73, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[24] Munawwar K.V, and Nandhini K, “Mitigating Class Imbalance in Offensive Language Detection in Malayalam Through NLPAUG,” International Journal of Engineering Research and Sustainable Technologies, vol. 2, no. 1, pp. 29-35, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[25] Navya K et al., “Detecting Homophobic and Transphobic Comments on Social Media in Malayalam and English Languages,” Procedia Computer Science, vol. 258 pp. 2479-2489, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[26] Abhinav Patil et al., “Multimodal Sentiment Analysis of Tamil and Malayalam,” Third Workshop on Speech and Language Technologies for Dravidian Languages, Varna, Bulgaria, pp.  250-330, 2023.
[Google Scholar]

[27] Sarbajeet Pattanaik, Ashok Yadav, and Vrijendra Singh, “Dll5143@DravidianLangTech 2025: Majority Voting-based Framework for Misogyny Meme Detection in Tamil and Malayalam,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 191-199, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[28] Premjith B et al., “Findings of the Shared task on Multimodal Sentiment Analysis and troll meme Classification in Dravidian Languages,” Proceedings of the Second Workshop on Speech and Language Technologies for Dravidian Languages, Association for Computational Linguistics, Dublin, Ireland, pp 254-260, 2022.
[CrossRef] [Google Scholar] [Publisher Link]

[29] Rachana K et al., “Mucs@DravidianLangTech2023: Sentiment Analysis in Code-Mixed Tamil and Tulu Texts using fastText,” Proceedings of the Third Workshop on Speech and Language Technologies for Dravidian Languages, INCOMA Ltd., Shoumen, Bulgaria, pp 258-265, 2023.
[Google Scholar] [Publisher Link]

[30] M. Rahul, R.R. Rajeev, and S. Shine, “Social Media Sentiment Analysis for Malayalam,” International Journal of Computational Science and Engineering, vol. 6, no. 6, pp. 48-53, 2018.
[Google Scholar]

[31] Raj M. Rahul, and Dhanya S. Pankaj, “Social-sum-Mal: A Dataset for Abstractive text Summarization in Malayalam,” ACM Transactions on Asian and Low-Resource Language Information Processing, vol. 23, no. 11, pp. 1-20, 2024.
[Google Scholar]

[32] Pradeep Kumar Roy, and Abhinav Kumar, “Sentiment Analysis on Tamil Code-Mixed text using Bi-LSTM,” FIRE, pp 1044-1050, 2024.
[Google Scholar]

[33] Koyyalagunta Krishna Sampath, and M. Supriya, “Transformer based Sentiment Analysis on Code Mixed Data,” Procedia Computer Science, vol. 233, pp. 682-691, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[34] Poorvi Shetty, “Sarcasm Identification in Dravidian Languages Tamil and Malayalam,” FIRE, pp 240-248, 2023.
[Google Scholar]

[35] Soumya S, and Pramod K.V, “Sentiment Analysis of Malayalam Tweets using Machine Learning Techniques,” ICT Express, vol. 6, no. 4, pp. 300-305, 2020.
[CrossRef] [Google Scholar] [Publisher Link]

[36] Harish Vijay V et al., “Code Conquerors@DravidianLangTech 2025: Deep Learning Approach for Sentiment Analysis in Tamil and Tulu,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp. 254-258, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[37] Malliga Subramanian et al., “KEC_TECH_TITANS@DravidianLangTech 2025: Abusive text Detection in Tamil and Malayalam Comments using Machine Learning,” Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages, Association for Computational Linguistics, pp 259-263, 2025.
[CrossRef] [Google Scholar] [Publisher Link] 

[38] Malliga Subramanian et al., “Empowering Sentiment Analysis in Social Media: Enhancing Abusive Tamil Comment Classification using Transformer Models,” Journal of Big Data, vol. 12, no. 1, pp. 1-36, 2025.
[CrossRef] [Google Scholar] [Publisher Link]