Application of Natural Language Processing and Text Mining to Identify Patterns in Construction-Defect Litigation Cases

Abstract

Recently, construction-defect litigation has upsurged across the United States. Disputes arise due to a variety of reasons, and result in a range of negative impacts on construction projects, such as increased cost, delay, profit loss, and inconvenience. Although the majority of these disputes settle out of court, a public trail of legal records exists. Previous research has generally been limited to exploring a small subset of such cases based on restricted access to records and data. This ongoing research automates systematic exploration of construction-defect lawsuits in the public domain by using modern computational capabilities of natural language processing and text mining to conduct a comprehensive survey of legal cases over the last 10 years. The approach of this research is to use coded text mining to automatically identify and analyze thousands of publicly available construction-defect cases. To perform such research, the authors developed a program that trolls the national legal database, LexisNexis. Key contributions include the development of a model that can find the frequencies of keywords in the cases and apply a statistical algorithm called Latent Dirichlet Allocation (LDA) to identify important topics and themes in order to classify the case data. The research demonstrates new methods for exploring publicly available construction-defect cases. Major challenges are identified and discussed. As exploratory research, the findings are intended to inform and motivate future study, which may lead to identification of broad-based trends in construction-defect litigation.