SentKG-BERT: Integrating Sentiment Features and Knowledge Graphs with BERT for Mental Health Intent Classification
Megha Agarwal
Computer Science and Information Systems, Shri Ramswaroop Memorial University, Barabanki, UP, India
Author
📌 DOI: https://doi.org/10.63920/tjths.300926004
🔑 Keywords: Intent Classification, Mental Health Conversa-tional Systems, BERT, Sentiment Analysis, Knowledge Graph Em-beddings, Conversational Artificial Intelligence
📅 Publication Date: 10 September 2026
📜 License:
This work is licensed under a Creative Commons Attribution 4.0 International License
- Share — Copy and Redistribute the material
- Adapt — Remix, Transform, and build upon the material
- The licensor cannot revoke these freedoms as long as you follow the license terms.
Abstract:
Mental health conversational systems need proper intent detection in order to give a proper response to users in distress. The conventional intent classification models usually do not capture emotional signals of a context and domain knowledge at the same time. In order to resolve this shortcoming, this paper puts forward SentKG-BERT, a hybrid model that incorporates sentiment features and a mental-health knowledge graph in a BERT-based framework to achieve better intent classification. The suggested model makes use of the BERT-based contextual embeddings, sentiment analysis-based emotional polarity, and domain-specific knowledge graph structural semantic relations. A mental health conversational dataset was experimented on involving five intent classes namely Greeting, Anxiety, Depression, Gratitude, and Neutral. Findings indicate that SentKG-BERT performs much better than baseline models such as TF-IDF + SVM, BiLSTM, and standard BERT. The proposed method has an accuracy of 95.4% and F1-score of 95.0% with significant gains in the recognition of emotionally nuanced intents. These results indicate the usefulness of combining emotional and knowledge-based attributes to improve intent identification in mental health chatbots
Download Full PDF Paper
📖 How to Cite
Megha Agarwal (2026). SentKG-BERT: Integrating Sentiment Features and Knowledge Graphs with BERT for Mental Health Intent Classification. TEJAS J. Technol. Humanit. Sci., Vol. 05, Issue 03. https://doi.org/10.63920/tjths.300926004
📊 Article Metrics
References
[1] M. Saleem and J. Kim, “Intent-aware data augmentation by leveraging generative AI for stress detection in social media texts,” PeerJ Computer Science, vol. 10, 2024, doi: 10.7717/peerj-cs.2156.
[2] S. Ji, T. Zhang, L. Ansari, J. Fu, P. Tiwari, and E. Cambria, “MentalBERT: Publicly available pretrained language models for mental healthcare,” in Proceedings of the 2022 Language Resources and Evaluation Conference (LREC 2022), 2022, pp. 7184–7190.
[3] M. Kentour and J. Lu, “An investigation into the deep learning approach in sentiment analysis using graph-based theories,” PLoS ONE, vol. 16, no. 12, 2021, doi: 10.1371/journal.pone.0260761.
[4] T. Nijhawan, G. Attigeri, and T. Ananthakrishna, “Stress detection using natural language processing and machine learning over social interactions,” Journal of Big Data, vol. 9, no. 1, 2022, doi: 10.1186/s40537-022-00575-6.
[5] K. Sharma, R. Patel, and S. Iyer, “Knowledge graph enhanced aspect-level sentiment analysis,” arXiv preprint arXiv:2312.10048, 2023.
[6] S. K. Baberwal, N. A. Shelke, and K. Anwar, “Systematic review of recent advances in multimodal sentiment analysis,” Information Retrieval Journal, vol. 28, no. 1, 2025, doi: 10.1007/s10791-025-09727-7.
[7] A. S. Talaat, “Sentiment analysis classification system using hybrid BERT models,” Journal of Big Data, vol. 10, no. 1, 2023, doi: 10.1186/s40537-023-00781-w.
[8] B. Ogunleye, H. Sharma, and O. Shobayo, “Sentiment informed Sentence-BERT ensemble algorithm for depression detection,” Big Data and Cognitive Computing, vol. 8, no. 9, 2024, doi: 10.3390/bdcc8090112.
[9] S. Singh, S. B. Verma, V. Sharma, S. M. Tiwari, and A. Agrawal, “Federated learning approaches based on blockchain in smart environments,” in Proceedings of the 4th International Conference on Mathematical Modeling and Computational Science (ICMMCS 2025), S. Pal and Á. Rocha, Eds., Lecture Notes in Networks and Systems, vol. 1400. Cham, Switzerland: Springer, 2025, doi: 10.1007/978-3-031-91008-1_18.
[10] A. Agarwal, S. B. Verma, B. K. Gupta, and S. Singh, “Strengthening cloud computing security: A malware prevention and detection framework at the hypervisor level,” Journal of Information Assurance and Security, vol. 19, no. 5, pp. 180–196, 2025, doi: 10.2478/ias-2024-0013.