圖書資訊學系Tang, Muh-ChyunMuh-ChyunTangHuang, Kuo-TingKuo-TingHuang2012-10-172018-05-302012-10-172018-05-302007http://ntur.lib.ntu.edu.tw//handle/246246/242814San JoseIn this paper we propose a naturalistic methodology tailored to test two term suggesting tools designed to support exploratory search with PubMed. Terms co-occur with user’s query terms in the initial returned set are extracted and present to the user in two ways: one makes use of the MeSH tress structure; the other rank terms by their mutual information with the query. The user study adopts a quasiexperimental design, comparing the performance of the two interfaces with the baseline PubMed interface on subject searches. Of particular interest is the usefulness of the proposed interfaces for difference types of search tasks. One novel aspect of our methodology is the use of real user and real search tasks in naturalistic work setting, instead of using assigned tasks in a controlled experimental setting. The performance criteria will be users’ satisfaction with the search results and search experiences.7601628 bytesapplication/pdfen-USInformation Retrieval EvaluationUser studyexploratory search interfaceconference paperhttp://ntur.lib.ntu.edu.tw/bitstream/246246/242814/3/Exploring the interaction effects between search tasks and .pdf