Chapter

Literature-based Discovery

Volume 15 of the series Information Science and Knowledge Management pp 133-152

Literature-Based Knowledge Discovery using Natural Language Processing

  • D. HristovskiAffiliated withInstitute of Biomedical Informatics, Medical Faculty, University of Ljubljana
  • , C. FriedmanAffiliated withDepartment of Biomedical Informatics, Columbia University
  • , T. C. RindfleschAffiliated withNational Library of Medicine
  • , B. PeterlinAffiliated withDivision of medical genetics, UMC

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Literature-based discovery (LBD) is an emerging methodology for uncovering nonovert relationships in the online research literature. Making such relationships explicit supports hypothesis generation and discovery. Currently LBD systems depend exclusively on co-occurrence of words or concepts in target documents, regardless of whether relations actually exist between the words or concepts. We describe a method to enhance LBD through capture of semantic relations from the literature via use of natural language processing (NLP). This paper reports on an application of LBD that combines two NLP systems: BioMedLEE and SemRep, which are coupled with an LBD system called BITOLA. The two NLP systems complement each other to increase the types of information utilized by BITOLA. We also discuss issues associated with combining heterogeneous systems. Initial experiments suggest this approach can uncover new associations that were not possible using previous methods.