Text mining technique for Chinese written judgment of criminal case

Abstract

Text mining has become an effective tool for analyzing text documents in automated ways. Conceptually, clustering, classification and searching of legal documents to identify patterns in law corpora are of key interest since it aids law experts and police officers in their analyses. In this paper, we develop a document classification, clustering and search methodology based on neural network technology that helps law enforcement department to manage criminal written judgments more efficiently. In order to maintain a manageable number of independent Chinese keywords, we use term extraction scheme to select top-n keywords with the highest frequency as inputs of the Back-Propagation Network (BPN), and select seven criminal categories as target outputs of it. Related legal documents are automatically trained and tested by pre-trained neural network models. In addition, we use Self- Organizing Map (SOM) method to cluster criminal written judgments. The research shows that automatic classification and clustering modules classify and cluster legal documents with a very high accuracy. Finally, the search module which uses the previous results helps users find relevant written judgments of criminal cases. © 2010 Springer-Verlag.