Bangla News Trend Observation using LDA Based Topic Modeling

Abstract

Topic Modelling is an essential field of natural language processing (NLP) that can be considered as a type of statistical model for extracting the abstract topics that have occurred in a collection of documents. Bangla is among the most popular and used languages around the world and nowadays innumerable Bangla texts are generated through digital and social media. So the significance of extracting knowledge from these data is invaluable for various sectors. However, the number of works in this field is inadequate because of the lack of proper datasets, tools, and applications. Therefore, preparing a convenient dataset in Bangla can be a great help for topic modeling as well as for other NLP related research. In this paper, we have addressed some of those complications by creating a proper dataset. Also, we have demonstrated a method of observing the Bangla media trend by applying Latent Dirichlet Allocation (LDA) on newspaper articles. The result of our experiment suggests that the proposed method can be an admissible way of utilizing news media data to observe media trends overtime properly.