Comparison of Named Entity Recognition models based on Neural Network in Biomedical

摘要

The need of automated information extraction increases with the increase in biomedical text. Named entity recognition is one of the core tasks in automatic information extraction. Performing named entity recognition tasks are quite challenging in biomedical field. These challenges include limited availability of annotated datasets and misclassification of entities having multiple meanings. Many Neural Network and deep learning-based models are developed for overcoming these challenges and for increasing the performance of named entity recognition tasks. This paper compares different models based on neural network architecture. The performance of these models is compared on JNLPBA dataset. The results show that Long short-term memory - conditional random field model with Wiki PubMed-PMC embeddings has outperformed other models by achieving highest precision and F1-score. CollaboNet model achieves the highest recall. Further analysis is needed to explore and compare the tools for performing named entity tasks in biomedical field.