Comparing CNN and LSTM character-level embeddings in BiLSTM-CRF models for chemical and disease named entity recognition

TL;DR

This paper compares CNN and LSTM character-level embeddings in BiLSTM-CRF models for chemical and disease NER, finding similar accuracy but different training efficiencies.

Contribution

It provides a comparative analysis of CNN and LSTM character embeddings in NER models, highlighting their performance and computational trade-offs.

Findings

Both CNN and LSTM embeddings achieve state-of-the-art NER performance.

CNN embeddings are computationally more efficient than LSTM embeddings.

LSTM embeddings significantly increase training time compared to CNN embeddings.

Abstract

We compare the use of LSTM-based and CNN-based character-level word embeddings in BiLSTM-CRF models to approach chemical and disease named entity recognition (NER) tasks. Empirical results over the BioCreative V CDR corpus show that the use of either type of character-level word embeddings in conjunction with the BiLSTM-CRF models leads to comparable state-of-the-art performance. However, the models using CNN-based character-level word embeddings have a computational performance advantage, increasing training time over word-based models by 25% while the LSTM-based character-level word embeddings more than double the required training time.