HITextracter System for Chemical and Gene/Protein Entity Mention Recognition in Patents

In this paper, a hybrid system was proposed for chemical entity mention recognition (CEMP) and gene/protein related object recognition (GPRO) in BeCalm challenge. Firstly, five individual machine learning-based subsystems were developed to identify chemical and gene/protein related entity mentions, that is, a bidirectional LSTM (long-short term memory, a variant of recurrent neural network)-based subsystem without any manually-crafted feature, a bidirectional LSTM-based subsystem with some manually-crafted features, a bidirectional LSTM-based subsystem with orthographic features learning, a CRF (conditional random field)-based subsystem and a SSVM (structured support vector machine)-based subsystem. Then, an ensemble learning-based classifier was deployed to combine all the results predicted by above individual subsystems. Evaluation on the official test set showed that the best F1-scores achieved by our system are 90.37% on CEMP, 76.34% on CPRO type 1 respectively.