Prototype of classifier for the decision support system of legal documents

Abstract

We propose a prototype of the classifier of electronic documents for the decision support system in the field of economic justice. The system uses both well-known text analytics algorithms and an original algorithm based on an artificial neural network. A text mining model has been developed to classify court documents to determine the category (class) of a statement of claim. A preliminary analysis of court documents and the selection of significant features were carried out. To choose the best way of solving problem of document classification we implemented Bayesian classification algorithm, k nearest neighbor algorithm and decision trees algorithm. All used algorithms show results with errors on the same sample corpus of texts. To improve the accuracy of classification, an original model based on an artificial neural network was developed, which shows an unmistakable determination of the type of document on a test sample for a number of classes of lawsuits in arbitration proceedings.