Guided Semi-Supervised Non-negative Matrix Factorization on Legal Documents

TL;DR

This paper introduces GSSNMF, a novel semi-supervised matrix factorization method that enhances classification and topic modeling of legal documents by integrating labels and seed words, outperforming previous methods.

Contribution

The paper presents GSSNMF, a new guided semi-supervised non-negative matrix factorization technique that jointly improves classification accuracy and topic coherence in legal document analysis.

Findings

GSSNMF outperforms SSNMF and Guided NMF in classification accuracy.

GSSNMF achieves higher topic coherence.

Application to legal documents demonstrates practical effectiveness.

Abstract

Classification and topic modeling are popular techniques in machine learning that extract information from large-scale datasets. By incorporating a priori information such as labels or important features, methods have been developed to perform classification and topic modeling tasks; however, most methods that can perform both do not allow for guidance of the topics or features. In this paper, we propose a method, namely Guided Semi-Supervised Non-negative Matrix Factorization (GSSNMF), that performs both classification and topic modeling by incorporating supervision from both pre-assigned document class labels and user-designed seed words. We test the performance of this method through its application to legal documents provided by the California Innocence Project, a nonprofit that works to free innocent convicted persons and reform the justice system. The results show that our proposed method improves both classification accuracy and topic coherence in comparison to past methods like Semi-Supervised Non-negative Matrix Factorization (SSNMF) and Guided Non-negative Matrix Factorization (Guided NMF).