The project focuses on developing techniques to increase the context window of encoder-based transformer models, such as RoBERTa, to improve the accuracy of entity extraction from lengthy legal documents. This project addresses the limitations of standard models with fixed context lengths, which struggle with long contracts, by implementing advanced methods like positional embedding interpolation and custom attention mechanisms. The goal is to achieve higher precision and recall in identifying legal entities, ensuring more reliable and comprehensive contract analysis.
Lithurshan Kanagalingam
A final year Computer Science and Engineering undergraduate deeply passionate about Machine Learning and web application development, I am committed to continuous learning and the adoption of emerging technologies, striving to enhance my expertise and deliver effective solutions.
Brief Description of Project
Recent News
- All
