Using Machine Learning to Support Quality Judgments

摘要

Resourceandcollection"quality"isbecominganincreasinglyimportanttopicforeducationaldigitallibraries.Computationalmodelsofqualityandautomatedapproachesforcomputingthequalityofdigitalresourcesarenecessarycomponentsofnextgenerationcognitivetoolsaimedatsupportingcollectioncuratorsinmakingqualitydecisions.Thisresearchidentifiesandcomputesmetricsfor16qualityindicators(e.g.,cognitiveauthority,resourcecurrency,cost,andadvertising)andemploysmachine-learningtechniquestoclassifyresourcesintodifferentqualitybandsbasedontheseindicators.Twoexperimentswereconductedtodetermineiftheseindicatorscouldbeusedtoaccuratelyclassifyresourcesintodifferentqualitybandsandtodeterminewhichindicatorspositivelyornegativelyinfluencedresourceclassification.Theresultssuggestthatresourcescanbeautomaticallyclassifiedintoqualitybands,andthatfocusingonasubsetoftheidentifiedindicatorscanincreaseclassificationaccuracy.