ISDB: Interaction Sentence Database

The Interaction Sentence Database (ISDB) allows users to retrieve sets of sentences fitting specified characteristics. To support this, a database of sentences from abstracts in MEDLINE was created. The sentences in the database all contain at least two biomolecule terms and one interaction-indicating term. A web interface to the database allows the user to query for sentences containing an interaction-indicating term, a single biomolecule name, or two biomolecule names, as well as for a list of biomolecules co-occurring with a given biomolecule in at least one sentence.The system supports researchers needing conveniently available sets of sample sentences for corpus-based research on sentence properties. It also illustrates a model architecture for a sentence-based retrieval system which would be useful to people seeking information and knowledge on-line. ISDB can be freely accessed over the Web at http://bioinformatics.ualr.edu/cgi-bin/services/ISDB/isdb.cgi webcite, and the processed database will be provided upon request.Traditionally, information obtained for biological research is stored as text in journals. When a paper is submitted to a journal it is often left to curators to pull information from the text into specialized resources for researchers [1]. Such specialized resources include interaction, biomolecule complex and pathway databases. As an alternative to reading a stack of papers, a simple query of a specialized database can then potentially return the information of interest. The development of better automated literature mining tools is essential if we are to take full advantage of the daunting amount of available information.Information extraction is the process of pulling facts from text [2]. Accurate software tools for information extraction are needed to automate the process of pulling these facts from text to populate specialized databases. The fragmentation of scientific expertise and the resulting highly specialized fields leads to disconne