Changes

2011talks Submissions

1,508 bytes added, 19:14, 12 November 2010
no edit summary
Claremont needed to perform a mass evaluation of item level records to facilitate large scale collection moves and de-accession. Our de-accession criteria, for example, include that 3 or more copies of any book must be available in the 50+ libraries in our Link+ network. We addressed our requirements with the help of the OCLC Worldcat Search and xID APIs and a couple simple python scripts. The process was ultimately a success. We will present our approach, code, and the lessons learned as we discovered limits inherent in the APIs and in our own coding (in)experience. Bonus sub-topic: the use of OCLC Work ID to identify and coalesce alternative ISBNs.
 
 
== A Simple Algorithm for User Query Classification & Resource Recommendation ==
* Josh Bishoff, University of Illinois, bishoff2 at illinois dot edu
 
One of the longstanding problems in library services is how we might automatically direct users to the most appropriate personnel, databases or facilities to meet their information need. Utilizing the faceted navigation features of various next-gen catalogs, we can efficiently & very accurately assign subject domains to user search queries.
 
For example: if a user searches “Gallium Arsenide” in the library discovery layer, we can first broadcast this query to a suitably large OPAC and receive the following subject distribution:
 
-Engineering & Technology: 45%
-Physical Sciences: 21%
-Education: 9%
…and so on.
 
By leveraging the cataloging efforts that have classified large collections, we can efficiently classify queries with a high rate of accuracy. By applying this approach to the library discovery layer, we can offer users tailored result sets from subject-specific A & I services. We can also recommend subject specialists & most appropriate campus libraries.
This presentation will discuss the technical challenges of implementing such a system, the trouble with mapping traditional subject classifications to non-book resources. The dangers of ''incorrect'' automatic query classification will be discussed, along with strategies to combat this. A functional system will be demonstrated and code will be made available.
Anonymous user