Nonlinear Dimensionality Reduction with Judicial Document Learning

This paper investigates the applications of NLP and machine learning techniques to judicial decision making.These legal documents are often represented by n-grams, term frequency-inverse document frequency (TF-IDF) or other methods, which lead to high feature representation of documents.Often, the number of labeled judicial documents are less than the dimensionality of features of judicial documents. It will degrade the prediction performance by directly using these extracted features from text. This paper studies the applications of various linear and non-linear dimensionality reduction techniques for judicial decision making. The extensive empirical experiments have been carried out to evaluate the manifold learning based dimensionality reduction method for judicial documents classification.