Files in this item
|(no description provided)|
|Title:||Retrieval Performance in a Full Text Journal Article Database|
|Department / Program:||Library Science|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||It is often assumed that the availability of the full text of journal articles for online searching will make abstracting and indexing obsolete and that search results with full text will be superior to the results of bibliographic searching. This study tested these assumptions by comparing results from searching the full text of journal articles in the Harvard Business Review Online database on a commercial search system with the results from searching on the titles and the value-added fields of abstract and controlled vocabulary.
There was a significant difference at the .05 level between the full text and other methods for total documents retrieved, recall ratio, and overlap and uniqueness of specific documents retrieved. Full text searching contributed a significantly greater number of documents, had a significantly higher recall, and contributed a significantly higher number of unique documents. Its precision ratio was lower than other search methods, however, and the cost per relevant document retrieved by full text was over twice that of the other methods.
Word occurrence patterns in the full text were shown to provide an aid in improving the precision ratio of full text searching. If a search word occurs frequently in a document or in more than four paragraphs of a document, that document is more likely to be relevant than would be expected by the average precision for all documents retrieved. Documents retrieved by both full text and controlled vocabulary searches are more likely to be relevant.
Titles contributed no relevant documents; whereas a combination search of controlled vocabulary, abstract, and full text was often necessary for comprehensive retrieval.
The results show that full text does provide the most complete retrieval, but in most cases it alone does not provide comprehensive retrieval. Controlled vocabulary terms and abstracts contribute unique documents and provide more cost effective searching if comprehensive retrieval is not needed.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1984.
|Date Available in IDEALS:||2014-12-15|