TheWorldWideWebcontinues to challenge organizations to makeonline access to their expertise convenient for their clients. Onemeansof expertise access that manyclients find convenientin everydaylife is askingnatural languagequestions of the organization. Tosupport it online, wedevelopedan approach to building organization-embeddedquestion-answering intermediaries, called InformationExchangesystems. Thesesystemsuse their knowledgeof the organization's structure to answerthe clients' questionsandto acquire newexpertise fromthe organization'sexperts. Ourapproachuses techniquesof hierarchical and predictive , combinedterm weighting, abstraction-based retrieval, and negativeevidenceacquisition. Weillustrate our approachwith the ChicagoInformationExchange system, an Information Exchangeapplication embeddedin one university's computerscience department.
Vladimir A. Kulyukin, Kristian J. Hammond, Robin D