한국어 법률 텍스트 처리 방안 연구 - 유사 판례 도출 태스크를 중심으로 -
A Study on Korean Legal Text Processing Techniques - Focusing on the task of deriving similar legal cases -
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
본 연구는 법률 시장에서의 유사 판결문 매칭 데이터셋의 수요가 증가함에도 불구하고 이에 대한 국내 연구가 부족함을 인식하고, 법률 문서를 사용하여 자연어 처리에 근본적으로 활용할 수 있는 한국어 법률 텍스트 특화 모델을 개발하기 위한 기초 실험을 진행하였다. 한국어 법률 텍스트 처리를 위해 세 가지 사전 학습된 모델을 미세 조정하였고 유사 판결 요지 추출 태스크를 수행하였다. 성능 평가를 위해 타겟 판결 요지와 후보 판결 요지 간의 유사도를 계산하였으며, 유사도를 바탕으로 추출된 판결 요지가 실제 법률 전문가와 일반 언어학 전공자의 직관에 부합하는지 판단하기 위해 정성적 평가를 진행하였다. 그 결과 법률 전문가가 일반 언어학 전공자에 비해 판결 요지 간 유사도를 낮게 평가하였으며 이를 통해 법률 전문가의 법률 텍스트 유사성 판단 기준이 모델과 일반 언어학 전공자와는 다르다는 것을 알 수 있었다. 최종 연구 결과로는 언어학 분야와 법률 분야의 학문적 융합이 필수적인 동시에 전문가 자문을 기반으로 한 한국어 법률 AI 모델 개발의 필요성을 확인하였다.
Abstract
Recognizing the lack of domestic research on similar case matching datasets in the legal market, this study conducted a basic experiment to develop a Korean legal text-specific model that can be fundamentally used in natural language processing using legal cases. For the Korean legal text processing, three different pre-trained models were fine-tuned and their performance was evaluated. In order to evaluate the performance of the model on whether the target summary and summary candidates extracted through the model are actually similar, the similarity between the summaries of the cases were measured and qualitative evaluation was conducted. As a result, legal experts evaluated the similarity between the summaries of cases lower than those majoring in general linguistics without legal expertise. It was confirmed that this is because the criteria for legal experts to judge the similarity of legal texts are different from those of general linguistics majors. Through this study, the necessity of developing a Korean legal AI model based on expert advice was confirmed. Future research will examine in-depth the Korean legal AI model that combines linguistics and legal fields as well as the construction of a Korean legal benchmark dataset for legal task processing.
참고문헌 (21)
- 김범준, "미국의 법률체계와 사법제도", 최신외국법제정보, 4, pp.181~206 (2010)
- Abhay Shukla, "Legal Case Document Summarization: Extractive and Abstractive Methods and their Evaluation", AACL, 2022, pp.2022(1), 1048~1064
- Bhattacharya, Paheli, "Methods for Computing Legal Document Similarity: A Comparative Study" (2020)
- Chaojun Xiao, "Cail2019-scm: A dataset of similar case matching in legal domain" (2019)
- Deepali Jain, "Summarization of legal documents: Where are we now and the way forward", Computer Science Review, 40 (2021)
- Hwang, W., "A Multi-Task Benchmark for Korean Legal Language Understanding and Judgement Prediction" (2022)
- Ilias Chalkidis, "LexGLUE: A Benchmark Dataset for Legal Language Understanding in English", the Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (2022)
- J. Devlin, "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding", NAACL-HLT, (1), pp.4171~4186 (2019), Association for Computational Linguistics
- J. Yoon, "Abstractive Summarization of Korean Legal Cases using Pre-trained Language Models", 2022 16th International Conference on Ubiquitous Information Management and Communication(IMCOM), pp.1~7 (2022)
- Le, Tho, "Unsupervised keyword extraction for japanese legal documents", Frontiers in Artificial Intelligence and Applications, 259, pp.97~106 (2013)
- Mandal, Arpan, "Measuring Similarity among Legal Court Case Documents", pp.1~9 (2017)[DOI]
- Martín Chozas, Patricia, "Challenges of terminology extraction from legal Spanish corpora", Proceedings of the 2nd Workshop on Technologies for Regulatory Compliance, pp.73~83 (2018)
- Nguyen, HT., "Transformer-Based Approaches for Legal Text Processing", Rev Socionetwork Strat, 16, pp.135~155 (2022)
- Quoc Le, "Distributed Representations of Sentences and Documents", ICML'14 Proceedings of the 31st International Conference on International Conference on Machine Learning, 32(9), pp.1188~1196 (2014)
- Sanh, Victor, "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter", ~ (2019)
- Kumar, S., "Similarity analysis of legal judgments", Bangalore Compute Conf (2011)
- Xiao, Chaojun, "CAIL2018: A Large-Scale Legal Dataset for Judgment Prediction" (2018)
- Xingyi Duan, "Cjrc: A reliable human-annotated benchmark dataset for chinese judicial reading comprehension", Proceedings of CCL (2019), Springer
- Cail2019-fe. Technical report, "Yi Shu", Gridsum, ~ (2019)
- Yoshinobu Kano, "Coliee-2018: Evaluation of the competition on legal information extraction and entailment", Proceedings of JSAI, pp.177~192 (2018), Springer
- Zhong, H., "How Does NLP Benefit Legal System: A Summary of Legal Artificial Intelligence" (2020)[DOI]