Kaviyarasu S


2024

pdf bib
Integration of Self-Attention Model with Intralingual Word Embedding for Contextual Semantic Analysis of Thirukkural Text
Shanthi Murugan | Kaviyarasu S | Balasundaram S R
Proceedings of the 21st International Conference on Natural Language Processing (ICON)

Thirukkural, one of the ancient works of Tamil Literature, is popular worldwide due to the moral values and practices it teaches to the society. Understanding the verses with meaning, especially context, is important. In this regard, this paper introduces a system designed to generate contextualized word meanings for the couplets of the Thirukkural, tailored to assist school children in understanding the text more effectively. Unlike traditional methods that provide detailed explanations in paragraph form, our method focuses on word-by-word interpretation, based on context through an integrated self-attention model. By combining the self-attention mechanism with FastText embeddings, our approach achieves improved performance over state-of-the-art models such as Word2Vec and standalone FastText. We evaluate the semantic understanding of the Thirukkural text using metrics as manual scoring. Tamil Thirukkural Agarathi serves as the gold-standard dataset for evaluation, demonstrating the effectiveness of our approach in capturing the nuanced semantics of the Thirukkural.