Generalized AutoNLP model for name entity recognition task

Abstract

Unsupervised pre-trained word embeddings have been widely used in recent studies in the field of Natural Language Processing. After the remarkable achievement obtained by the introduction of BERT in various NLP related tasks, studies had been more focused on deep-learning based approach to represent the raw input sequence of string words.

However, there is an uncertainty of these deep-learning based approaches able to convey all the semantic meanings of words and have generalized ability on AutoNLP on name entity recognition related tasks. In this project, we have proposed an architecture of a combination of deep-learning based approach word embeddings, BERT with static word embeddings, GloVe. Experiments are conducted to study the performance of our proposed architecture with BERT word embeddings on AutoNLP name entity recognition tasks.