While modern digital libraries and their information retrieval systems are based on search engine technology, they rank their search results by means of fairly simple algorithms. They are still relying on text-matching and the weighing of metadata fields. Opportunities provided by multi-dimensional ranking functions have not been taken yet.
The insufficiency of these retrieval systems becomes more obvious if we consider users which are heavily affected by their daily use of popular search engines and the social web. They simply judge the ranking by the quality of the single results and the compilation of well-balances result sets. Moreover, they expect those results to be listed first, which are popular among similar users.
Against this background, the research project LibRank – funded by the German Research Foundation (DFG) – targets at the development of more sophisticated and multi-dimensional ranking algorithms for bibliographical information systems. By exploiting insights from web search behavior, the project explores how
(a) relevance can be determined in terms of factors like freshness, popularity or availability of items,
(b) the usage context influences the relevance of search results, and
(c) search results can be delivered and rendered more conform to user expectations derived from web search behavior.