A Study on Knowledge Extraction from Official Bangla Documents

Abstract

Bangla is the seventh largest spoken language in the world. However, the information searching in the digital Bangla papers is a tiresome job as its ends up with incorrect and a very few information. It is difficult because wide computational resources for Bangla are very limited. It is literally infeasible to list and analyze the Bangla data manually. Several approaches for identifying and extracting tables, figures, emotion, reviews, and algorithms have been done in English. Furthermore, knowledge extraction in Bangla documents for emotion or opinion detection and sentence extraction for summarization have been explored. However, they do not provide enough textual information for the user for Bangla text content. In this work, we proposed a domain specific composite approaches to find out the agendas, and its decisions from the minutes of meeting of academic council of Khulna University of Engineering & Technology (KUET) with Query-based features, Content-based features, Context-based features and Semantic features. We also demonstrated the techniques to categorize the knowledge where a single query is given by the user and displayed the result sequentially by date. All the operations are presented with sufficient theoretical analysis and experimental results.