DeepHCoref: A Deep Neural Coreference Resolution for Hindi Text

Abstract

Coreference Resolution is the process of detecting a cluster of mentions that point to the same entity. The Coreference Resolution will enhance the performance of numerous Natural Language Processing (NLP) applications viz. Machine Translation, Question Answering, Chatbots, Text Summarization, etc. The state-of-the-art Coreference Resolution model for Hindi is based on a Rule-based and machine learning approach. This paper presents the Coreference Resolution system for Hindi based on Bi-GRU-CNN and Biaffine classifier with IndicBERT and MuRIL BERT. According to our knowledge, the first Coreference resolution system based on deep learning is developed, particularly for Hindi. The motivation behind this is the scarcity of resources available for Hindi and to diminish the dominion of hand-crafted features used by the previous Coreference resolution model in the Hindi language and present a new state-of-the-art Coreference Resolution model for Hindi. The coreference annotated dataset is used for the Hindi language, containing 3.6K verbalizations and 78K tokens from the news article domain. There is a lot of work done in this field for English, but minimal amount of work is presented in Hindi. The experimental results received are commendable regarding the Precision, Recall, and F-measure.