Estimating Legal Document Structure by Considering Style Information and Table of Contents

Abstract

Text analytics is used to analyze diverse documents. For example, legal documents (such as contracts, ordinances, regulations, and global standards) must be analyzed for corporations to manage their business risk and meet compliance requirements. However, since documents are often stored or published as documents without a common structure, they need to be preprocessed to analyze them in subsequent text analytics. In particular, the following two forms of preprocessing are useful for text analytics: (1) extracting text, and (2) estimating document structure (such as chapters, sections, and subsections), which is used to define the range of topics or articles in a document. This paper presents a preprocessing method to estimate document structure from documents without a common structure. The proposed method follows rule-based approach, and consists of three algorithms: (1) one is based on style information, such as bold font; (2) another is based on numbered objects, such as sections; and (3) the other is based on a document’s Table of Contents, which summarizes the document’s structure. The accuracy of the proposed method is also evaluated by using 102 documents. The proposed method was found to be able to estimate document structure with 96.6% accuracy.