An approach to improve kernel-based Protein-Protein Interaction extraction by learning from large-scale network data

Protein-Protein Interaction extraction (PPIe) from biomedical literatures is an important task in biomedical text mining and has achieved desirable results on the annotated datasets. However, the traditional machine learning methods on PPIe suffer badly from vocabulary gap and data sparseness, which weakens classification performance. In this work, an approach capturing external information from the web-based data is introduced to address these problems and boost the existing methods. The approach involves three kinds of word representation techniques: distributed representation, vector clustering and Brown clusters. Experimental results show that our method outperforms the state-of-the-art methods on five publicly available corpora. Our code and data are available at: http://chaoslog.com/improving-kernel-based-protein-protein-interaction-extraction-by-unsupervised-word-representation-codes-and-data.html.