Deep Named Entity Recognition in Hindi Using Neural Networks

Abstract

In Natural Language Processing, named entity recognition (NER) is a task of (NLP) that tries to automatically identify and annotate Named Entities in text, such as people, places, and organisations. We employ a deep learning-based architecture in this work to solve the problem of recognising named entities in a Hindi text phrase. In the literature, approaches based on bidirectional long short-term memory (BiLSTM) have been utilized for the NER task. We performed recursive BiLSTM in this study, which includes a de-noising autoencoder with conditioning logic. Experiments were undertaken to examine the behaviour of individual word embeddings and batch sizes, which is vital for training deep models. The findings of the suggested system architecture, as well as a comparison of performance characteristics with existing systems, are presented in this study.