Named Entity Recognition of Protein Complex Incorporating Syntactic Information

Abstract

In order to automatically identify protein complex named entities from unstructured biomedical literature, we propose a protein complex named entity recognition method which incorporates syntactic information. Through incorporating three types of syntactic information, including part-of-speech labels, syntactic constituents and dependency relations with a key-value memory network, the model is able to learn syntactic information features, thus solving the problem of utilizing the syntactic information insufficiently. On the data set of protein complex constructed by distant supervision method, the experimental results show that after incorporating the three syntactic information with the key-value memory network, the models outperform the original model in different degrees, indicating that our proposed method can effectively improve the performance of protein complex named entity recognition model.