Researchers have developed a novel framework to perform topic modeling on trilingual parliamentary debates from Sri Lanka, encompassing Sinhala, Tamil, and English. This system overcomes challenges posed by complex PDF layouts, multilingual scripts, and agglutinative morphology by employing LLM-based text extraction and a multilingual embedding and clustering pipeline. The BiTopic extension further enhances interpretability and noise reduction. Applied to over 19,000 speeches from 2017-2026, the framework successfully identified 30 macro-topics, aligning with significant national events like the 2019 Easter Sunday attacks and the 2022 economic crisis, outperforming traditional LDA methods. AI
IMPACT Enables deeper analysis of multilingual political discourse and historical events through advanced NLP techniques.
RANK_REASON The cluster describes a research paper published on arXiv detailing a new computational method for analyzing multilingual text data. [lever_c_demoted from research: ic=1 ai=1.0]
- 2022 economic crisis
- arXiv
- Bitopic protein
- Easter Sunday attacks
- English
- Hugging Face
- LDA
- Sinhala
- Sri Lanka
- Tamil
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
