Text summarization of legal documentation with explainability feature

Teh, Xiao Thong (2024) Text summarization of legal documentation with explainability feature. Project Report. Melaka, Malaysia, Universiti Teknikal Malaysia Melaka. (Submitted)

Text (Full text) Text summarization of legal documentation with explainability feature.pdf - Submitted Version Download (16MB)

Abstract

Summarizing lengthy documents, especially in the legal domain, poses significant challenges for both humans and automated systems. Human efforts often entail considerable time and effort while automated systems sometimes falter in decision-making, leading to ambiguity in the generated summaries. This project explores the use of text summarization in legal documentation, coupled with an explainability feature. It addresses the challenges of condensing lengthy legal texts and improving transparency in automated summarization systems. The project involves gathering legal documents, developing a BART summarization model, and integrating explainability within the system, which is visualizing attention mechanism. The system which has been deployed on web-based application is the final product of this project. The system performance, that includes BERTScore, cosine similarity, and ROUGE score between human-generated and system-generated summaries, and evaluation by target users, lead to several engaging insights on legal summarization. The model demonstrated moderate performance where the user feedback indicated satisfaction with its functionality but highlighted the need for user interface improvements. Key strengths of the model include the system’s explainability that is crucial for legal applications. Future improvements suggested including refining model training, enhancing the user interface, and adding features like adjustable summary lengths and language translation. The project contributes valuable insights to artificial intelligence and natural language processing with potential for further research and development.

Actions (login required)

View Item