Bidirectional LSTM-CNNs with Extended Features for Named Entity Recognition

Abstract

Named Entity Recognition (NER) is vital preprocessing step for many Natural Language Processing applications such as relation extraction and question answering. NER has been studied for decades. While, much of the earlier studies for NER have focused on using powerful features and knowledge resources, recent studies using Deep Learning techniques have not needed to use these powerful features and knowledge resources. Instead, approach of these studies is learning powerful features from the data by itself. In this paper, we extend a bidirectional LSTM-CNN model by adding syntactic and word-level features. Evaluation shows that adding syntactic and semantic information about words without feature engineering to the model surpasses the baseline model tested on CoNLL-2003 English NER dataset.