Nepali POS Tagging using Deep Learning Approaches

EasyChair Preprint 2073

Abstract

Part of Speech (POS) tagging is one of the fundamental task in Natural Language Processing (NLP). It plays vital role in various NLP applications such as machines translation, text-to-speech conversion, question answering, speech recognition, word sense disambiguation and information retrieval. It is also referred as grammatical tagging or word-category disambiguation which is a process of labeling every word in sentences with tag based on its context and syntax of the language. It is challenging to develop promising POS tagger for morphologically rich language like Nepali. This paper focuses on implementing and comparing different deep learning based POS tagger for Nepali such as Simple Recurrent Neural Network (RNN), Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bi-directional Long Short Term Memory (Bi-LSTM).  These approaches were trained and tested in a corpus of Nepali tag set. The result shows that Bi-directional LSTM outperforms all other three approaches.

Keyphrases: Natural Language Processing, Recurrent Neural Network, part-of-speech

@booklet{EasyChair:2073,
  author    = {Sarbin Sayami and Tej Bahadur Shahi and Subarna Shakya},
  title     = {Nepali POS Tagging using Deep Learning Approaches},
  howpublished = {EasyChair Preprint 2073},
  year      = {EasyChair, 2019}}