ResearchTeam_HCN at SemEval-2023 Task 6: A knowledge enhanced transformers based legal NLP system

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 1245–1253 July 13-14, 2023 ©2023 Association for Computational Linguistics ResearchTeam_HCN at SemEval-2023 Task 6: A knowledge enhanced transformers based legal NLP system Dhanachandra Ningthoujam†, Pinal Patel∗, Rajkamal Kareddula∗, Ramanand Vangipuram† ∗RAAPID.ai †HEALTHCARENLP SOFTECH LLP †{dhanachandra.n,ramanandh.v}@healthcarenlp.com ∗{pinal,rajkamal}@raapid.ai Abstract This paper presents our work on LegalEval (understanding legal text), one of the tasks in SemEval-2023. It comprises of three sub-tasks namely Rhetorical Roles (RR), Legal Named Entity Recognition (L-NER), and Court Judge- ment Prediction with Explanation (CJPE). We developed different deep-learning models for each sub-tasks. For RR, we developed a multi- task learning model with contextual sequential sentence classification as the main task and non- contextual single sentence prediction as the sec- ondary task. Our model achieved an F1-score of 76.50% on the unseen test set, and we at- tained the 14th position on the leaderboard. For the L-NER problem, we have designed a hybrid model, consisting of a multi-stage knowledge transfer learning framework and a rule-based system. This model achieved an F1-score of 91.20% on the blind test set and attained the top position on the final leaderboard. Finally, for the CJPE task, we used a hierarchical ap- proach and could get around 66.67% F1-score on judgment prediction and 45.83% F1-score on the explainability of the CJPE task, and we attained 8th position on the leaderboard for this sub-task. 1 Introduction The court pending cases are rapidly raising in pop- ulous countries like India. Towards this challenge, systems that can automatically process, understand and organize legal documents need to be designed. Another major challenge of processing legal doc- uments is that they are unstructured, very lengthy, contain a lot of legal jargon, etc. Thus, the exist- ing NLP systems can not be directly applied. The SemEval-2023 (Modi et al., 2023) provides 3 differ- ent problems on the Indian legal documents namely: 1. Rhetorical Roles (RR) (Kalamkar et al., 2022b) (Malik et al., 2021a): Segmenting a legal document into semantically similar sentences. The RR task can be defined as the classification of sentences in a legal document into pre-defined classes. 2. Le- gal Named Entity Recognition (L-NER) (Kalamkar et al., 2022a): Legal entity extraction from the legal document. The task is to find out the legal entities like Court names, Petitioner names, Respondents, Judge names, and case numbers mentioned in a given legal document. 3. Court Judgement Predic- tion with Explanation (CJPE)(Malik et al., 2021b): Court judgment output prediction with explana- tions. Given a legal document, the task is to find out the final outcome of the case i.e whether the appeal raised by the petitioner is accepted or not. In case of acceptance, generate the explanation of the case. The datasets for all the above sub-tasks are available in the English language. In this work, we present our proposed solutions for the above three problems. We proposed a multi-task learning neu- ral network model to solve the sub-task RR. Our model consists of two classification layers. The first classifier …

Content truncated to 500 words