Louvain Modularity and Ensemble Gradient Boost Absolute Congruence Deep Belief Classifier for Legal Documents

Abstract

As far as the Indian law system is concerned, judgment is contemplated as a final decision that is stated by the court for a stipulated case. Over the past few years, owing to the mushrooming growth in technologies, like, machine learning, deep learning and ensemble learning, judgments are kept in digital forms for ease of use. With the emergence of an automated system, the proceedings are made simple for the law professional. Several law firms have already started applying these learning mechanisms that specifically concentrate on emphasizing information from the Indian judgmental document. Also extracting key factors from legal documents is considered as a time consuming process. To ease the task of classifying legal documents, a method called Louvain Modularity and Ensemble Gradient Absolute Congruence (LM-EGAC) deep belief classifier for legal documents is proposed. The LM-EGAC method is split into two sections, namely optimized topic modeling (i.e., extracting optimized topic words) and classification (i.e., identifying case type). First, Highly Associative Unique Keyword and Louvain Modularity-based Optimized Topic Modeling are applied for extracting optimal topic words for further processing. Followed by which the Ensemble Gradient Absolute Congruence boosting model is applied for classifying the legal documents into corresponding cases. The Ensemble Gradient boosting algorithm initially forms an Absolute Congruence function with the purpose of classifying input legal documents along with the optimized topic words. Next, the results of the deep belief classifier are joined to construct a strong classifying by minimizing the error with the aid of binary entropy function and Louvain Modular function. The proposed LM-EGAC method is experimentally validated with several performance factors like, precision, recall, error and computational time. The experimental results quantitatively confirm that the proposed LM-EGAC method achieves better precision with minimum error and computation time upon comparison with the traditional methods.