Legal Information as a Complex Network: Improving Topic Modeling through Homophily

Natural language processing is a key component to computational legal science. Network analysis is also very important to further understand the structure of references in legal documents. In this paper, we improve topic modeling for legal documents by using homophily derived from two family of references: prior cases and statute laws. We propose a new analysis of a rich data set in order to create these networks. The use of the reference-induced homophily topic modeling improves on prior methods.