Identification of latent legal knowledge in differing site condition (DSC) litigations

Conflicts in construction projects have always been a major problem. Unless an alternate resolution mechanism is spelled out in the contract, these disputes are typically resolved in court, which might be time consuming and financially substantial. This paper represents a continuation in a research focused on creating robust methodologies for legal decision support within the construction industry. Consequently, this papers tackles the problem of automating the extraction of implicit knowledge about significant legal factors upon which verdicts of Differing Site Condition (DSC) litigations are based. To that end, the research methodology (1) utilized a set of 600 cases from the Federal Court of New York; (2) adopted 15 legal concepts that have been found to be statistically significant for DSC litigations; (3) implemented 4 weighing mechanism for data representation, namely Term Frequency, Logarithmic Term Frequency, Augmented Term Frequency, and Term Frequency Inverse Document Frequency; and (4) employed Machine Learning (ML) classifiers, namely Naïve Bayes, Decision Tree, and PART for the development of 12 prediction models. Among the finding of this study (1) ML classifiers present a suitable solution for the analyzed task; and (2) Naïve Bayes classifiers achieved the highest prediction accuracy of 88%.