Learning Logical Structures of Paragraphs in Legal Articles

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 20–28, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Learning Logical Structures of Paragraphs in Legal Articles Ngo Xuan Bach, Nguyen Le Minh, Tran Thi Oanh, Akira Shimazu School of Information Science, Japan Advanced Institute of Science and Technology, 1-1 Asahidai, Nomi, Ishikawa, 923-1292, Japan {bachnx,nguyenml,oanhtt,shimazu}@jaist.ac.jp Abstract This paper presents a new task, learning logical structures of paragraphs in legal ar- ticles, which is studied in research on Le- gal Engineering (Katayama, 2007). The goals of this task are recognizing logi- cal parts of law sentences in a paragraph, and then grouping related logical parts into some logical structures of formulas, which describe logical relations between logical parts. We present a two-phase framework to learn logical structures of paragraphs in legal articles. In the first phase, we model the problem of recog- nizing logical parts in law sentences as a multi-layer sequence learning problem, and present a CRF-based model to recog- nize them. In the second phase, we pro- pose a graph-based method to group logi- cal parts into logical structures. We con- sider the problem of finding a subset of complete sub-graphs in a weighted-edge complete graph, where each node corre- sponds to a logical part, and a complete sub-graph corresponds to a logical struc- ture. We also present an integer linear pro- gramming formulation for this optimiza- tion problem. Our models achieve 74.37% in recognizing logical parts, 79.59% in recognizing logical structures, and 55.73% in the whole task on the Japanese National Pension Law corpus. 1 Introduction Legal Engineering (Katayama, 2007) is a new re- search field which aims to achieve a trustworthy electronic society. Legal Engineering regards that laws are a kind of software for our society. Specif- ically, laws such as pension law are specifications for information systems such as pension systems. To achieve a trustworthy society, laws need to be verified about their consistency and contradiction. Legal texts have some specific characteristics that make them different from other kinds of doc- uments. One of the most important characteristics is that legal texts usually have some specific struc- tures at both sentence and paragraph levels. At the sentence level, a law sentence can roughly be di- vided into two logical parts: requisite part and ef- fectuation part (Bach, 2011a; Bach et al., 2011b; Tanaka eta al., 1993). At the paragraph level, a paragraph usually contains a main sentence1 and one or more subordinate sentences (Takano et al., 2010). Analyzing logical structures of legal texts is an important task in Legal Engineering. The outputs of this task will be beneficial to people in under- standing legal texts. They can easily understand 1) what does a law sentence say? 2) what cases in which the law sentence can be applied? and 3) what subjects are related to the provision de- scribed in the law sentence? This task is the pre- liminary step, which supports other tasks in legal text processing (translating legal articles into …

Content truncated to 500 words