Identification of POS Tags for the Khasi Language based on Brill’s Transformation Rule-Based Tagger

Abstract

Khasi is a Mon-Khmer language that belongs to the Austro-Asiatic language family. Khasi language is spoken by the indigenous people of the state Meghalaya in the North-Eastern part of India. The main purposeof this paper is to develop Part-of-Speech (PoS) tagger for the Khasi language using a Rule-based approach. To work on POS tagging, one needs a grammatically tagged corpus. However, the Khasi language does not have a standard corpus for PoS tagging. Therefore, another aim or purpose of this paper is to develop a Khasi lexicon or POS corpus and using the Rule-Based Brill’s Transformation to automatically tag the given Khasi text. While anticipating the challenges in building such acorpus, this paper has brought out an analysis based on the Khasi corpus of around 1,03,998 words in its initial phase. We also show in this paper how the Khasi corpusis created. By using Brill’s Transformation rule-based learning on the created corpus in this preliminary study, accuracies of 97.73% and 95.52% were obtained on validating data and testing data respectively. This workis the first attempt to investigate POS tagging using the rule-based model with the designed Khasi POS corpus.