Personalized Search based on a User-centered Recommender Engine Leyla Zhuhadar and Olfa Nasraoui Knowledge Discovery and Web Mining Lab Dept. of Computer Engineering and Computer Science University of Louisville, Louisville, KY 40292, USA leyla.zhuhadar@wku.edu, olfa.nasraoui@louisville.edu Abstract—Designing personalized search engines based on a recommender system that takes into consideration the user situated moment in relation to the subject matter and the context that governs user interest has been largely ignored. In this paper, we present a novel approach to integrating user interests into search within a recommender system that is guided by the semantic representation of the user and the content. In addition, our research tackles two problems of creating any recommender system: (a) the initial stage problem (how to provide recommendations to a user if the system hasn’t been used yet) and (b) user context (providing the same user with different recommendations based on the context of their recent activity). Also the design of our recommender system is modular. It integrates and accommodates user’s preferences by using User Relevance Feedback. Finally, we describe how the personalization aspects can increase the recommendation quality. Index Terms—semantic; recommender search engine; ontol- ogy; user relevance feedback I. I NTRODUCTION In this paper we present a novel approach to integrating user semantic interests within a recommender system for a personalized search engine. First, to overcome the problem of the initial stage (a.k.a cold start) of designing any recom- mender system (how to provide recommendations to a user if the system has not been used yet), our system generates an initial profile for the user at the moment the user logs in to the platform based on the similarity between the user semantic profile and the E-learning domain ontologies. During the registration process when the user selects the department and the course(s) he/she is interested in, based on these selections, an initial ontology profile is created for this particular user. Therefore, semantically related topics (courses/lectures) are automatically recommended while the user is searching for resources. Second, while more similar users are registered into the system, over a period of time, the recommendation schema dynamically changes based on the learners’ activities and the similarity between the users’ semantic profiles. Also, we provide the user with the capability to change his/her recommendations preferences at any moment using a User Relevance Feedback technique which provides the user with the capability to prune his/her ontology profile by excluding documents he/she is not interested in. II. BACKGROUND AND RELATED WORK As we described in our introduction, our proposed E- learning personalized search engine recommends learning ob- jects to a user based on his/her initial or evolving semantic profile. There are several related works that tackle the problem of recommending resources to users in both: the industry and academia domains. The most popular recommender search engines in industry are: (1) Amazon.com: recommendations based on similar items (items-based recommendations); it uses item-to-item collaborative filtering; this algorithm matches each of the user’s purchased and rated items to similar items, then combines those similar items into a recommendation list. To find the most similar match for a given item, the recommender engine first has to build a similar-items table by finding items that customers tend to purchase together [1], and (2) Google Personalized News (news.google.com): recom- mendations are based on the similarity between user profiles; this results in user-based recommendations,. This is one of the most scalable recommender systems that provides personalized news for millions of subscribers; Google News uses three types of collaborative filtering techniques: (1) MinHash Clustering, (2) Probabilistic Latent Semantic Indexing (PLSI), and (3) Covisitation counts [2]. Recently, recommendation systems have become the center of attention to many researchers in E-learning. [3] presented an automated way to find reviewers’ interest from the Web without the need for asking the review- ers to list their key interests, then, distributing conferences papers accordingly. [4] implemented an E-learning system that provides students with recommendations for technical papers. The platform is an adaptive platform to both user and open Web materials. Therefore, resources might be added or deleted based on the user ratings interests of previously visited papers. In general, this system shares with our platform the ability to model and handle an evolving user as well as evolving resources. However, it does not tackle the problem of the initial stage or cold start of recommendations. Moreover, it recommends technical papers based on user’s rating to previ- ously visited documents. As we know, most learners would not spend time in rating resources, especially, in a setting that involves studying. Users might delete some resources that they found them uninteresting, but will not rate each visited 2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology 978-0-7695-4191-4/10 $26.00 © 2010 IEEE DOI 10.1109/WI-IAT.2010.296 200