U-CREAT: Unsupervised Case Retrieval using Events extrAcTion

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 13899–13915 July 9-14, 2023 ©2023 Association for Computational Linguistics U-CREAT: Unsupervised Case Retrieval using Events extrAcTion Abhinav Joshi∗ Akshat Sharma∗ Sai Kiran Tanikella∗ Indian Institute of Technology Kanpur (IIT Kanpur)Ashutosh Modi {ajoshi, akshatsh, tskiran, ashutoshm}@cse.iitk.ac.in Abstract The task of Prior Case Retrieval (PCR) in the legal domain is about automatically citing rel- evant (based on facts and precedence) prior legal cases in a given query case. To further promote research in PCR, in this paper, we pro- pose a new large benchmark (in English) for the PCR task: IL-PCR (Indian Legal Prior Case Retrieval) corpus. Given the complex nature of case relevance and the long size of legal documents, BM25 remains a strong baseline for ranking the cited prior documents. In this work, we explore the role of events in legal case retrieval and propose an unsupervised re- trieval method-based pipeline U-CREAT (Un- supervised Case Retrieval using Events Extrac- tion). We find that the proposed unsupervised retrieval method significantly increases perfor- mance compared to BM25 and makes retrieval faster by a considerable margin, making it appli- cable to real-time case retrieval systems. Our proposed system is generic, we show that it generalizes across two different legal systems (Indian and Canadian), and it shows state-of- the-art performance on the benchmarks for both the legal systems (IL-PCR and COLIEE cor- pora). 1 Introduction Traditionally, in the legal domain, for a given legal case (query document) at hand, lawyers and judges have relied on their expertise and experience to cite relevant past precedents (cited documents). More- over, even when legal professionals have made lim- ited use of technology, it has been mainly restricted to Boolean queries and keywords. However, as cases increase, it becomes difficult for even experi- enced legal professionals to cite older precedents. NLP-based technologies can aid legal profession- als in this regard. The task of Prior Case Retrieval (PCR) has been formulated to address this problem (Rabelo et al., 2022). More concretely, the task of ∗Equal Contributions These statements were forwarded to the Police DET NOUN AUX VERB ADP DET NOUN det nsubjpass auxpass prep pobj det (statement, forward, police) NOUN VERB NOUN Extracted Event Figure 1: Dependency parse of the sentence (along with extracted event) from the IL-PCR corpus: “These statements were forwarded to the Police". <NAME> open door it contain particular body send hospital body send examination <NAME> conduct body body send hospital body send examination <NAME> conduct autopsy <NAME> conduct body injury find body there be injury <ORG> conduct <NAME> <NAME> conduct body injury find body there be injury <ORG> conduct <NAME> it not have book conspiracy raid police station police station attack article there be evidence who identify appellant evidence run scene he ransack house box take number "<NAME> open door" , "body send examination", "<NAME> conduct autopsy", "there be injury", "injury find body", "there be evidence" Candidate Doc Events Query Doc Events Matching Events <NAME> sit room <NAME> …

Content truncated to 500 words