Paper: Identifying Interaction Sentences from Biological Literature Using Automatically Extracted Patterns

ACL ID W09-1317
Title Identifying Interaction Sentences from Biological Literature Using Automatically Extracted Patterns
Venue Workshop on Biomedical Natural Language Processing
Session
Year 2009
Authors

An important task in information retrieval is to identifysentencesthatcontainimportantrelation- ships between key concepts. In this work, we proposea novel approachto automaticallyextract sentencepatternsthat containinteractionsinvolv- ing concepts of molecular biology. A pattern is definedin this work as a sequenceof specialized Part-of-Speech(POS) tags that capturethe struc- ture of key sentences in the scientific literature. Eachcandidatesentencefor the classificationtask is encoded as a POS array and then aligned to a collectionof pre-extracted patterns. The qual- ity of the alignment is expressed as a pairwise alignmentscore. The mostinnovative component of this work is the use of a Genetic Algorithm (GA) to maximizethe classificationperformance of the alignment scoring scheme. The sy...