Application of Topic Models to Judgments from Public Procurement Domain
This work addresses the need for better information retrieval and analysis in legal texts for domain experts, but it is incremental as it applies existing methods to a new domain.
The researchers tackled the problem of automatically analyzing themes in a large corpus of public procurement judgments using latent Dirichlet allocation (LDA) combined with unsupervised keyword extraction, resulting in improved interpretability and computational performance for detecting recurring themes and temporal trends in contract appeals.
In this work, automatic analysis of themes contained in a large corpora of judgments from public procurement domain is performed. The employed technique is unsupervised latent Dirichlet allocation (LDA). In addition, it is proposed, to use LDA in conjunction with recently developed method of unsupervised keyword extraction. Such an approach improves the interpretability of the automatically obtained topics and allows for better computational performance. The described analysis illustrates a potential of the method in detecting recurring themes and discovering temporal trends in lodged contract appeals. These results may be in future applied to improve information retrieval from repositories of legal texts or as auxiliary material for legal analyses carried out by human experts.