Re-ranking for personalization using concept hierarchy in DL environment

Abstract

In Digital Library (DL) system, users interact with the system to search for books or research papers. Users can search through metadata or search for information in the pages by querying using keywords. In both cases, a huge amount of results are returned; however, the relevant ones to the user are not often amongst the top few. Re-ranking of the search results based on the user's interest has received wide attention in information retrieval. This work presents extending conventional search engine to searching digital library data of user's interests. The proposed system improve information access by building knowledge about a user, acquired using the user's interaction with the system, in order to customize information access. User profiling is done using a hybrid approach by taking into consideration login details and click-through data. This system mapping framework automatically maps Dmoz Open Directory Project (ODP) topics to users' interests and takes advantage of manually edited data available in ODP, to categorize and personalize search results, according to user interests. Reranking of the results is done based on user interests. This makes it easy to find relevant DL pages faster than normal search engines. Performance has been evaluated for online DL systems. System's performance improves by 16.55% if the average per query is calculated and approximately 10% if per user average is calculated over baseline (Google CSE) after re-ranking.