ItemConference Paper

Emotion Trajectory Analysis and Model Comparison for Hate Speech and Radicalization Detection in Code-Mixed Platforms

Title
Emotion Trajectory Analysis and Model Comparison for Hate Speech and Radicalization Detection in Code-Mixed Platforms
Creator
Sharanya, R.; Singh, Shivangi; Jose, Deepa V
Description
The growing presence of multilingual and codemixed content on social media creates major challenges for automated emotion recognition and mental health support. In this work, we introduce an emotion-aware computational framework that processes code-mixed Indian language comments and predicts user emotions with high accuracy, followed by context-aware support suggestions. Our dataset comes from the AI4Bharat IndicNLP corpus [14] and the Dravidian-CodeMix sentiment dataset [15], featuring a variety of multilingual user comments. To maintain linguistic consistency, we translate the raw texts into English using Google Translator and then preprocess them through normalization, tokenization, and stopword removal. We use three advanced transformer-based models, DistilBERT (six emotions), DistilRoBERTa (seven emotions), and RoBERTa GoEmotions (27+ emotions), to categorize the emotions in the comments. We compare predictions across the models and select the most reliable label for each text, which is further verified through manual checks with human annotators. This process results in a curated dataset labeled with emotions and enriched with model provenance. With this dataset, we train a Logistic Regression classifier using TF-IDF features to create an efficient, explainable prediction pipeline. The system classifies emotions and provides tailored suggestions based on emotional states, improving user support in online interactions. Experimental results show the robustness of the pipeline and its ability to adapt to various code-mixed inputs. This study offers an integrated dataset-model-suggestion framework that advances emotion recognition in multilingual contexts and supports the creation of practical emotion-aware digital systems. 2025 IEEE.
Source
2025 17th IEEE International Conference on Computational Intelligence and Communication Networks, CICN 2025;pp.1378-1383
Date
01-01-2025
Publisher
Institute of Electrical and Electronics Engineers Inc.
Subject
Code-Mixed Social Media; Computational Social Science; Early Detection; Emotion Dynamics; Hate Speech Detection; Natural Language Processing (NLP); Radicalization; Temporal Emotion Trajectories
Coverage
Sharanya R., Deemed to Be University, Dept. of Computer Science Christ, Bangalore, India; Singh S., Deemed to Be University, Dept. of Computer Science Christ, Bangalore, India; Jose D.V., Deemed to Be University, Dept. of Computer Science Christ, Bangalore, India
Rights
Restricted Access; Hardcopy may be available in the library
Relation
ISBN: 979-833158733-8;
Format
online
Language
English
Type
Conference paper

Citation

Sharanya, R.; Singh, Shivangi; Jose, Deepa V, “Emotion Trajectory Analysis and Model Comparison for Hate Speech and Radicalization Detection in Code-Mixed Platforms,” CHRIST (Deemed To Be University) Institutional Repository, accessed June 18, 2026, https://archives.christuniversity.in/items/show/25786.

Output Formats