NotesFAQContact Us
Search Tips
ERIC Number: ED314059
Record Type: Non-Journal
Publication Date: 1988-Jun
Pages: 124
Abstractor: N/A
Reference Count: N/A
Extracting Knowledge for Intermediary Expert Systems: The Selection of Search Keys. Final Report.
Fidel, Raya
This study investigated online searching behavior manifested by 39 experienced professional searchers performing their regular, job-related searches in order to uncover the rules they use for the selection of search keys, and to represent these rules in a formal model that could be used in the construction of intermediary expert systems. The case study method with controlled comparison was used, and data analyses were based on two existing models: the Selection Routine, a decision tree presenting the rules used to select search keys by eight searchers in a previous study, and a list of moves, or modifications, in search strategies that was based on observations of the searching behavior of the same eight subjects. Two types of moves were identified: operational moves that preserve the meaning of a request, and conceptual moves that change the meaning of a request. Within each type, the moves are presented in three groups: precision moves; recall moves; and moves to increase precision and recall. Data analysis involved measuring the frequency with which each type of search key was selected, each move was selected, and a reason was cited to explain the selection of a search key. The statistical associations among 11 variables were also examined for each search: (1) number of search keys selected; (2) percentage of free-text terms used; (3) frequency with which a thesaurus was not consulted; (4) number of databases used; (5) number of moves made; (6) percentage of operational moves made; (7) number of precision moves made; (8) number of recall moves made; (9) percentage of recall moves made; (10) the searcher's subject area specialty; and (11) the searcher's work environment. It was concluded that searching behavior is lawful and follows certain patterns; current systems cannot provide satisfactory recall and some are even an impediment to useful searching practices; and the requests presented by users are the least predictable of the factors that affect searching behavior. (6 tables, 3 figures, and 42 references) (SD)
Publication Type: Reports - Research; Tests/Questionnaires
Education Level: N/A
Audience: N/A
Language: English
Sponsor: National Science Foundation. Washington, DC. Div. of Information Science and Technology.
Authoring Institution: Washington Univ., Seattle. Graduate School of Library and Information Science.