Problem classification of legal bar exam and structured dictionary design for human relationships in legal documents

Abstract

Legal document processing is one of promising applications of natural language processing. However, legal sentences are very complicated and require various difficult analysis techniques. We focus on the competition workshop COLIEE (Competition for Legal Information Extraction and Entailment) that aims to automatically answer the legal bar exam. Firstly, we manually classify the past legal bar exam data distributed in COLIEE by required analysis techniques such as conditional sentence extraction and negative form interpretation. Secondly, we focus on person relationships among our classifications, propose a dictionary design for legal document processing based on verbs and their arguments with roles of persons.