Parse Tree Database for Information Extraction

Information extraction systems are traditionally implemented as a pipeline of special-purpose processing modules targeting the extraction of a particular kind of information. A major drawback of such approach is that whenever a new extraction goal emerges or a module is improved, extraction has to be re-applied from scratch to the entire text corpus even though only a small part of the corpus might be affected. In this paper, we describe a novel approach for information extraction so that extraction needs are expressed in the form of database queries, which are evaluated and optimized by databases. Using database queries for information extraction enables generic extraction and minimizes reprocessing of data. In addition, our approach provides two different query generation components that can automatically form database queries for extraction from training datasets, as well as from unlabeled data through a mechanism inspired by the pseudo-relevance feedback approach found in protein-protein interactions and drug-protein-metabolic relations from two sets of corpus. Experiments show that our approach achieves a precision of 83.6% and recall of 58.6% (F-measure of 64.2%) for the extraction of protein-protein interactions from the BioCreative 2 corpus, while achieving a precision of 85.0% and recall of 26.0% (F-measure of 39.8%) for drug-protein-metabolic relations.