CSECU-DSG at SemEval-2023 Task 6: Segmenting Legal Documents into Rhetorical Roles via Fine-tuned Transformer Architecture

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 2112–2117 July 13-14, 2023 ©2023 Association for Computational Linguistics CSECU-DSG at SemEval-2023 Task 6: Segmenting Legal Documents into Rhetorical Roles via Fine-tuned Transformer Architecture Fareen Tasneem, Tashin Hossain, Jannatun Naim, and Abu Nowshed Chy Department of Computer Science and Engineering University of Chittagong, Chattogram-4331, Bangladesh {fareen.tasneem,tashin.hossain.cu, jannatun.naim.cu}@gmail.com, and nowshed@cu.ac.bd Abstract Automated processing of legal documents is essential to manage the enormous volume of le- gal corpus and to make it easily accessible to a broad spectrum of people. But due to the amor- phous and variable nature of legal documents, it is very challenging to directly proceed with complicated processes such as summarization, analysis, and query. Segmenting the documents as per the rhetorical roles can aid and acceler- ate such procedures. This paper describes our participation in SemEval-2023 task 6: Sub-task A: Rhetorical Roles Prediction. We utilize a finetuned Legal-BERT to address this task. We also conduct an error analysis to illustrate the shortcomings of our deployed approach. 1 Introduction Analyzing and extracting critical information from legal documents poses a great challenge for re- searchers. Legal documents can vary substantially in content, format, and structure depending on the jurisdiction, domain, and nature of documents. The corpus of legal documents is expanding rapidly, resulting in the need for automated systems for the effective retrieval of consequential information. Moreover, to ensure that these legal documents are utilized efficiently and justly, they must be accessi- ble to a wide range of people. The task benefiting from the automation of legal documents are case law analysis, summarization, semantic search, and so on. Identifying the rhetorical role of each seg- ment in the legal documents can aid in these tasks. Rhetorical role identification refers to partitioning the document into semantically meaningful seg- ments with each segment having specific roles and purposes. These rhetorical roles can be used to ex- tract significant information for further processing and analysis. But due to legal documents being unstructured, varying from domain to domain, and The first three authors have equal contributions. containing domain-specific literature, the identifi- cation of rhetorical roles is a challenging task. To address this challenge, SemEval-2023 task 6 (Modi et al., 2023) incorporated Sub-task A: Rhetorical Roles Prediction (RR). The task is to segment a given legal document by predicting the rhetorical role label for each sentence. An instance of a docu- ment segmented according to the rhetorical roles is illustrated in Figure 1. Hachey and Grover (2004) employed a feature set consisting of the location of the sentence in the document, thematic words, sentence length, and cue phrases. They utilized decision tree, naive bayes, winnow algorithm, and support vector ma- chine (SVM) as the classifiers. On the other hand, Saravanan et al. (2008) investigated rule-based learning algorithms (such as SLIPPER) and con- ditional random fields (CRF) along with features like indicator features, local features, layout fea- tures, state transition features, and legal vocabulary features. A multitask learning (MTL) framework incorporating Bi-LSTM and CRF …

Content truncated to 500 words