This proposal advocates a continuation of the MEDLINE/Full-Text Research Project which has studied aspects of the retrieval of relevant articles from full-text databases. In carrying out this phase of the project we shall make extensive use of the data which we have accumulated in answering previous research questions. Two of the four new questions to be addressed arise directly from the results of the current quest for heuristics for full-text searching. They are: 1) How many times should a searcher request that a given concept appear in a document--and in which fields--in order to insure optimum retrieval? 2) What effect does the ability to search both the indexing and the natural language of full-text have on retrieval? The third question results from the increasing use of the structured abstract in medical journals; it is, 3) Does the full-text retain its recall advantage now that the structured abstract is available in the bibliographic files? This question arises from the possibility that the structured abstract might be sufficient for retrieval since it covers certain variables which would be of importance in a search. The fourth question has been suggested by physician participants in the project and focusses our search for heuristics on specific types of relevant articles. It is: 4) How can the searcher of full-text manipulate the strategy to retrieve only relevant articles of highest quality? A corollary questions arises: if quality filters are applied to retrieval from full-text databases does the advantage of full-text over MEDLINE remain? The advent of new searching software for full-text in both online systems and CD-ROM products now makes it possible to investigate the first two questions. Data accumulated so far suggests that there is a close relationship between kind of concept being searched and the most effective search heuristic. By repeating a search in both MEDLINE with the structured abstract and in a full-text journal file-we should be able to see if the full-text retains its recall advantage. Finally the obvious need for """"""""quality filters"""""""" in medical information retrieval is the impetus for a search for heuristics for the retrieval of the """"""""best"""""""" in full-text medical literature.

Agency
National Institute of Health (NIH)
Institute
National Library of Medicine (NLM)
Type
Research Project (R01)
Project #
5R01LM004605-08
Application #
2237633
Study Section
Special Emphasis Panel (SRC)
Project Start
1993-09-01
Project End
1997-08-31
Budget Start
1995-09-01
Budget End
1996-08-31
Support Year
8
Fiscal Year
1995
Total Cost
Indirect Cost
Name
University of Missouri-Columbia
Department
Biostatistics & Other Math Sci
Type
Schools of Medicine
DUNS #
112205955
City
Columbia
State
MO
Country
United States
Zip Code
65211
Johnson, E D; McKinin, E J; Sievert, M E et al. (1997) An analysis of objective quality indicators on Year Book citations: implications for MEDLINE searchers. Bull Med Libr Assoc 85:378-84
Sievert, M E; McKinin, E J; Johnson, E D et al. (1996) Beyond relevance--characteristics of key papers for clinicians: an exploratory study in an academic setting. Bull Med Libr Assoc 84:351-8
Johnson, E D; Sievert, M C; McKinin, E J (1995) Retrieving research studies: a comparison of bibliographic and full-text versions of the New England Journal of Medicine. Proc Annu Symp Comput Appl Med Care :846-50
McKinin, E J; Sievert, M E; Collins, B R (1991) Currency of full-text medical journals: CCML and MEDIS vs. MEDLINE. Bull Med Libr Assoc 79:282-7
Sievert, M C; McKinin, E J; Johnson, E D et al. (1991) Retrieval from full-text medical literature: the dream & the reality. Proc Annu Symp Comput Appl Med Care :348-52
McKinin, E J; Sievert, M; Johnson, E D et al. (1991) The Medline/full-text research project. J Am Soc Inf Sci 42:297-307