Biomedical Relationship Extraction from Literature Based on Bio-semantic Token Subsequences

Relationship Extraction (RE) from biomedical literature is an important and challenging problem in both text mining and bioinformatics. Although various approaches have been proposed to extract protein-protein interaction types, their accuracy rates leave a large room for further exploration of more effective methods. In this paper, two supervised learning algorithms based on newly-defined “bio-semantic token subsequence” are proposed for multi-class biomedical relationship extraction. The first approach calculates a “bio-semantic token subsequence kernel”, while the second one explicitly extracts weighted features from bio-semantic token subsequences. The proposed structure called “bio-semantic token subsequence” is able to capture semantic features from natural language sentences for biomedical RE. Two supervised learning algorithms based on the proposed structure outperform the state-of-the-art biomedical RE methods on multi-class protein-protein interaction extraction.