Viettel-AI at SemEval-2023 Task 6: Legal Document Understanding with Longformer for Court Judgment Prediction with Explanation

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 862–868 July 13-14, 2023 ©2023 Association for Computational Linguistics Viettel-AI at SemEval-2023 Task 6: Legal Document Understanding with Longformer for Court Judgment Prediction with Explanation Thanh Dat Hoang, Chi Minh Bui, and Khac-Hoai Nam Bui Viettel Cyberspace Center, Viettel Group, Viet Nam {datht17,minhbc,nambkh}@viettel.com.vn Abstract Court Judgement Prediction with Explanation (CJPE) is a task in the field of legal analysis and evaluation, which involves predicting the out- come of a court case based on the available le- gal text and providing a detailed explanation of the prediction. This is an important task in the legal system as it can aid in decision-making and improve the efficiency of the court process. In this paper, we present a new approach to understanding legal texts, which are normally long documents, based on data-oriented meth- ods. Specifically, we first try to exploit the char- acteristic of data to understand the legal texts. The output is then used to train the model using the Longformer architecture. Regarding the ex- periment, the proposed method is evaluated on the sub-task CJPE of the SemEval-2023 Task 6. Accordingly, our method achieves top 1 and top 2 on the classification task and explanation task, respectively. Furthermore, we present several open research issues for further investigations in order to improve the performance in this re- search field. 1 Introduction The sub-task of CJPE invites participants to analyze a legal judgment document to predict the outcome of the case and justify their prediction with rele- vant sentences from the document that contributes to the decision(Malik et al., 2021b). The objec- tive of predicting court judgments is to enhance the efficiency and accuracy of the legal system by providing decision-makers with valuable pre- dictions and insights. Transformer-based models, such as BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019), currently dominate most natural language processing (NLP) tasks, including text classification. However, the quadratic complexity of their attention mechanisms imposes limits on the maximum input length (512 sub-word tokens in BERT, RoBERTa), which may not be able to handle all the meaning in the legal domain where longer documents are common(Malik et al., 2021b). To address this issue, sparse-attention models, such as LongFormer (Beltagy et al., 2020) and Big- Bird (Zaheer et al., 2020), increase the maximum input length to 4096 sub-word tokens, which may improve the performance in understanding legal documents. This paper focuses on investigating the data and experimenting with various large models, as we believe this sub-task involves processing long documents. Using the Longformer and TF-IDF features(Zhang et al., 2011), we achieve superior results and apply them to the explanation task. In the explanation task, we design a linear explana- tion algorithm and show the mechanism effectively inference the most important sentences leading to the decision of judgment. We also provide exten- sive experiments to analyse the capability of our proposed model on capturing semantic information in judgments. The main contributions of this study are summarized below: …

Content truncated to 500 words