Speaker
Dr
Thomas Hickey
(OCLC Research)
Description
One aspect of meeting users’ needs 'instantaneously' is immediate feedback, offering suggestions, or even returning results, while the user is typing.
We have experimented with both instant results and search suggestions in several environments. Recently we have implemented auto-suggest servers for searches of the FAST subject heading vocabulary and names in the Virtual International Authority File. As in searching, ranking is a crucial component of auto-suggest and we have explored a number of different approaches.
The auto-suggest service is derived from information collected during database indexing and is designed for the very fast response times (less than a tenth of a second) needed to support useful auto-suggest. We believe the techniques developed to be of general use and plan to make the code open source and port it into other search environments, such as Lucene.
Currently the auto-suggest responses are returned in JSON format. The contents of the response are driven by XSLT transformations of the underlying XML records, and therefore quite flexible. In addition to ‘traditional’ user search suggestions, we expect this approach to be extremely useful in metadata creation environments.
Thomas B Hickey & Ralph LeVan
OCLC Research
Primary author
Dr
Thomas Hickey
(OCLC Research)
Co-author
Mr
Ralph LeVan
(OCLC Research)
