Relevance Ranking in the Scholarly Domain
- Tamar Sadeh, PhD, Ex Libris Group, firstname.lastname@example.org
code4lib 2012, Tuesday 7 February 2012, 13:20-13:40
The greatest challenge for discovery systems is how to provide users with the most relevant search results, given the immense landscape of available content. In a manner that is similar to human interaction between two parties, in which each person adjusts to the other in tone, language, and subject matter, discovery systems would ideally be sophisticated and flexible enough to adjust their algorithms to individual users and each user’s information needs.
When evaluating the relevance of an item to a specific user in a specific context, relevance-ranking algorithms need to take into account, in addition to the degree to which the item matches the query, information that is not embodied in the item itself. Such information, which includes the item’s scholarly value, the type of search that the user is conducting (e.g., an exploratory search or a known-item search), and other factors, enables a discovery system to fulfill user expectations that have been shaped by experience with Web search engines.
The session will focus on the challenges of developing and evaluating relevance-ranking algorithms for the scholarly domain. Examples will be drawn mainly from the relevance-ranking technology deployed by the Ex Libris Primo discovery solution.