Topic modelling of legal documents using NLP and bidirectional encoder representations from transformers

Abstract

<span>Modeling legal text is a difficult task because of its unique features, such as lengthy texts, complex language structures, and technical terms. During the last decade, there has been a big rise in the number of legislative documents, which makes it hard for law professionals to keep up with legislation like analyzing judgements and implementing acts. The relevancy of topics is heavily influenced by the processing and presentation of legal documents in some contexts. The objective of this work is to understand the legal judgement corpus related to cases under the Hindu Marriage Act of India. The study looked into various methods to generate sentence embeddings from the judgement. This paper employs the power of the BERTopic algorithm for generating significant topics.</span>

Evaluating LDA and LSA for Topic Modeling in the Indonesian Natural Disaster

“…• Here is the parameter explanation [24] as shown in Table 1. There are many algorithms that can be used in analyzing topics with various characteristics, but the most popular one is LDA, LDA is an unsupervised learning algorithm that views documents as Bag of Words (BoW), LDA works by making key assuming, where LDA will select a set of words for each topic, the basic idea of topic modeling research is topics consisting of several relevant keywords which are then arranged into topics of documents [3], [5]- [7] as shown in Figure 3 .…”

“…Topic Modeling is one of the methods in text collection analysis or commonly called NLP (Natural Language Processing), this method is commonly employed for analyzing textual data gathered from databases or documents and can be applied to find topics based of set documents [2], [3]. This method will model relevant topics, in a collection of documents, articles or texts, modeling and visualizing topics relevant to natural disasters in Indonesia, will certainly make it easier to monitor and understand the current situation regarding natural disasters in Indonesia.…”

“…LDA assumes that every document comprises a blend of different subjects, and the presence of each word is linked to a particular topic. LSA, on the other hand, employs singular value decomposition to capture the latent semantic structure within a document-term matrix [3]- [7] .…”

Automated decision classification model for tax appeals commission in Morocco using latent dirichlet allocation

“…Topic modeling is a modern NLP technique that enables the exploration and prediction of hidden themes within large text datasets, regardless of the languages used [2], [3]. This method proves particularly valuable when dealing with extensive documents that are impractical to read and summarize manually.…”