A new method for automatic pattern acquisition to extract information from biomedical texts

Recently there have been many information extraction tasks applied to the biomedical domain, some of which contribute to extract protein-protein interactions from biomedical texts. This paper presents a new method for automatic pattern acquisition to extract protein interactions. The system automatically generates patterns by aligning sequences of tags of sentences from unlabeled corpus. To obtain a high tagging accuracy, we propose a morphology-based tagging method with a pre-tagging strategy for Brill's tagger. Our method differs from the previous pattern acquisition algorithms in the ways: first, it does not need to provide any seed word or pattern before the algorithm runs; second, we do not apply any parsing algorithm. Lastly, our method, which is based on dynamic programming, is fast.