Named Entity Recognizer for less resourced language Kokborok

摘要

Named Entity Recognition refers to the process of classifying text elements into predefined categories such as person names, organizations, locations, date, quantities etc. In this paper, we described the development of a rule based and a supervised Named Entity Recognizer for the Kokborok language which is less computerized and agglutinative. We used suffix information and Named Entity dictionary for the rule based system, while features like parts-of-speech (POS), context information and suffix etc. were used to develop the supervised system. Margin Infused Relaxed Machine Learning Algorithm is used for developing the supervised system. We achieved the maximum F-score of 83.18% after inclusion of the post-processing technique.