Current research of bio-text mining mainly focuses on event extractions. Biological networks present much richer and meaningful information to biologists than events. Bio-entity coreference resolution (CR) is a very important method to complete a bio-event’s attributes and interconnect events into bio-networks. Though general CR methods have been studies for a long time, they could not produce a practically useful result when applied to a special domain. Therefore, bio-entity CR needs attention to better assist biological network extraction. In this article, we present two methods for bio-entity CR. The first is a rule-based method, which creates a set of syntactic rules or semantic constraints for CR. It obtains a state-of-the-art performance (an F1-score of 62.0%) on the community supported dataset. We also present a machine learning-based method, which takes use of a recurrent neural network model, a long-short term memory network. It automatically learns global discriminative representations of all kinds of coreferences without hand-crafted features. The model outperforms the previously best machine leaning-based method.

Even a simple biological phenomenon may introduce a complex network of molecular interactions. Scientific literature is one of the trustful resources delivering knowledge of these networks. We propose LitWay, a system for extracting semantic relations from texts. LitWay utilizes a hybrid method that combines both a rule-based method and a machine learning-based method. It is tested on the SeeDev task of BioNLP-ST 2016, achieves the state-of-the-art performance with the F-score of 43.2%, ranking first of all participating teams. To further reveal the linguistic characteristics of each event, we test the system solely with syntactic rules or machine learning, and different combinations of two methods. We find that it is difficult for one method to achieve good performance for all semantic relation types due to the complication of bio-events in the literatures.

No abstract