Optimization ofParameters forEffective Web Information RetrievalUsinganEvolutionary
Brijesh Verma · 2005
Inthis paperwepresent anapproach based onthe application ofanevolutionary algorithm tooptimally tunethe parameters ofanovel technique foreffective webinformation retrieval. Context matching isa context-based technique for thead-hoc retrieval ofwebdocuments thatrelies onanumber ofinter-related parameters thatdefine thenatureofthecon- textituses. Itsaimistodynamically generate acontext-based measure oftermsignificance during retrieval thatcanbeused asanindicator ofdocument relevancy andultimately contrib- utetoadocuments rankscore. Buttheoptimal setting ofcon- textmatching parameters isanimportant aspect ofthetech- nique toensure effective retrieval. Thus, thegoalofthis paper istoinvestigate theuseofanevolutionary algorithm forthe optimization ofcontext matching parameters andcompareits performance toan iterative technique thatexhaustively ex- plores combinations ofparameters. We showhowthemost effective settings forparameters areobtained efficiently through theevolutionary algorithm. We also showhowcontext matching, through theuseoftheseoptimized parameters, achieves effective retrieval results onbenchmark datathatare asignificant improvement onpreviously published results.