Framework for Generation of Annotated Corpus for Recognition & Categorization of Geo-Political Entities from Indian Election Domain

This paper presents a comprehensive framework for generating an annotated corpus that facilitates the recognition and categorization of geo-political entities within the Indian Election domain. It focuses on the generation of an annotated corpus and the development of a customized Named Entity Recognition (NER) model for the identification and categorization of geo-political entities within Indian Election domain text. The entities of interest include states, districts, parliamentary constituencies, and assembly constituencies. The proposed work involves extensive research and data analysis from various resources such as ECI Publication manuals, several websites, Wikipedia, among others. The collected data is then annotated using the specialized annotation tool called Prodigy. Subsequently, a customized NER model is built and trained using the annotated data. The performance of the developed model is thoroughly tested and analyzed. The results of this project indicate the successful generation of a dataset specifically tailored for the Indian election domain, achieving an accuracy rate of 68%. Additionally, a framework is established for the generation of similar datasets in the future. It combines data analysis, annotation techniques, and machine learning to create a robust NER model that can effectively recognize and categorize geo-political entities in Indian Election domain texts. The generated dataset and framework contribute to the advancement of natural language processing techniques in the context of political analysis.