Sumanth Balaji, supervised by Prof. Vasudeva Varma received his Master of Science – Dual Degree in Computational Linguistics (CL). Here’s a summary of his research work on Legal text-recommendation systems:
Contracts are essential discourse units that are frequent in several day-to-day business workflows, especially between companies and governmental organisations. The sheer volume of textual contracts and the cumbersome nature of contract drafting and analysis allows for a strong case for applying Natural Language Processing (NLP) techniques to simplify tasks in legal NLP. The field of legal text recommendation is an emerging area of research that aims to automate the creation of legal documents and contracts through the recommendation of clauses for addition. However, the complexity of legal language and the need for precision and accuracy in legal documents make it a challenging task. Legal documents can be characterised by their high inter-sentence similarity and topic-specific content. For example, Zhong et al. (2020) showed that the sentences in legal corpora are almost 20% similar to each other. Drafting contracts by legal counsel is a manual process of taking a skeletal set of clauses and adding or modifying them for the contract goal. Generating such clauses with minimal user-provided information can be of significant benefit in automating legal contract drafting. This forms the main focus of the thesis- Towards text recommendation systems in legal domain documents with a focus on aided contract drafting. We investigate techniques for improving legal text recommendation systems. Additionally, this thesis also examines the use of interactive interfaces to enhance user engagement and interaction with the legal text generation system. The thesis first begins with our initial experiments and exploration of interactive recommendation systems in the legal domain, such as the Extractive Question answering pipeline and Rental agreement drafting. This is followed by an Investigation of strategies to improve Legal Contract Drafting of the current state of the art to discover optimizations as well as limitations. Afterward, we propose a controllable graph-based mechanism that can generate legal clauses using only the topic or type of legal clauses. Our pipeline consists of two stages: a graph-based planner and a clause generator. The planner outlines the content of a legal clause as a sequence of keywords in the order of generic to more specific clause information based on the input topic using a controllable graph-based mechanism. The generation stage takes in a given plan and generates a clause. We illustrate the effectiveness of our proposed two-stage approach on a broad set of clause topics in contracts.
July 2023