User profiles, descriptions of user interests, can be used by search engines to provide personalized search results. Many approaches to creating user profiles collect user information through proxy servers (to capture browsing histories) or desktop bots (to capture activities on a personal computer). Both these techniques require participation of the user to install the proxy server or the bot. In this study, we explore the use of a less-invasive means of gathering user information for personalized search. In particular, we build user profiles based on activity at the search site itself and study the use of these profiles to provide personalized search results. By implementing a wrapper around the Google search engine, we were able to collect information about individual user search activities. In particular, we collected the queries for which at least one search result was examined, and the snippets (titles and summaries) for each examined result. User profiles were created by classifying the collected information (queries or snippets) into concepts in a reference concept hierarchy. These profiles were

User profiling is an important part of the Semantic Web as it integrates the user into the concept of Web data with machine-readable semantics. In this paper, user profiling is presented as a way of providing the user with his/her interest-focused browsing history. We present a system that is incorporated into the Internet Explorer and maintains a dynamic user profile in a form of automatically constructed topic ontology. A subset of previously visited Web pages is associated with each topic in the ontology. By selecting a topic, the user can view the set of associated pages and choose to navigate to the page of his/her interest. Each topic can be seen as an interest of the user (hence the term interest-focused browsing history). The ontology is constructed by transforming the textual contents of the pages into sparse word-vectors and applying bisecting k-means clustering (i.e. a form of hierarchical clustering) on the set of sparse vectors. The most recently visited pages are used to identify the user’s current interest and map it to the ontology. The user can clearly see which topics, and their corresponding pages, are related (or are not related, for that matter) to his/her current interest. We see this as a useful way of organizing the user’s browsing history. To illustrate the functioning of the system, we demonstrate its behavior in one particular real-life scenario. 1

Personalized web search is a promising way to improve search quality by customizing search results for people with individual information goals. However, users are uncomfortable with exposing private preference information to search engines. On the other hand, privacy is not absolute, and often can be compromised if there is a gain in service or profitability to the user. Thus, a balance must be struck between search quality and privacy protection. This paper presents a scalable way for users to automatically build rich user profiles. These profiles summarize a user’s interests into a hierarchical organization according to specific interests. Two parameters for specifying privacy requirements are proposed to help the user to choose the content and degree of detail of the profile information that is exposed to the search engine. Experiments showed that the user profile improved search quality when compared to standard MSN rankings. More importantly, results verified our hypothesis that a significant improvement on search quality can be achieved by only sharing some higher-level user profile information, which is potentially less sensitive than detailed personal information.

Abstract. Web users use search engine to find useful information on the Internet. However current web search engines return answer to a query independent of specific user information need. Since web users with similar web behaviors tend to acquire similar information when they submit a same query, these unseen factors can be used to improve search result. In this paper we present an approach that mines these unseen factors from web logs to personalized web search. Our approach is based on probabilistic latent semantic analysis, a model based technique that is used to analyze co-occurrence data. Experimental results on real data collected by MSN search engine show the improvements over traditional web search. 1

An ontology can be seen as an explicit description of the concepts and relationships that exist in a domain. In this paper, we address the problem of building an interest ontology and predicting potential friendship relations between users in the social network Live Journal, using features constructed based on the interest ontology. Previous work has shown that the accuracy of predicting friendship links in this network is very low if simply interests common to two users are used as features and no network graph features are considered. Thus, our goal is to organize users ’ interests in an ontology (specifically, a concept hierarchy) and to use the semantics captured by this ontology to improve the performance of learning algorithms at predicting if two users can be friends. We have designed and implemented a hybrid clustering algorithm, which combines hierarchical agglomerative and divisive clustering paradigms, and automatically builds the interest ontology. Furthermore, we have explored the use of this ontology to construct interest-based features and shown that the resulting features improve the performance of various classifiers for predicting friendship links.

Context modeling has long been acknowledged as a key aspect in a wide variety of problem domains. In this paper we focus on the combination of contextualization and personalization methods to improve the performance of personalized information retrieval. The key aspects in our proposed approach are (1) the explicit distinction between historic user context and live user context, (2) the use of ontology-driven representations of the domain of discourse, as a common, enriched representational ground for content meaning, user interests, and contextual conditions, enabling the definition of effective means to relate the three of them, and (3) the introduction of fuzzy representations as an instrument to properly handle the uncertainty and imprecision involved in the automatic interpretation of meanings, user attention, and user wishes. Based on a formal grounding at the representational level, we propose methods for the automatic extraction of persistent semantic user preferences, and live, ad-hoc user interests, which are combined in order to improve the accuracy and reliability of personalization for retrieval.

Finding meaningful phrases in a document has been studied in various information retrieval systems in order to improve the performance. Many previous statistical phrase-finding methods had a different aim such as document classification. Some are hybridized with statistical and syntactic grammatical methods; others use correlation heuristics between words. We propose a new phrase-finding algorithm that adds correlated words one by one to the phrases found in the previous stage, maintaining high correlation within a phrase. Our results indicate that our algorithm finds more meaningful phrases than an existing algorithm. Furthermore, the previous algorithm could be improved by applying different correlation functions. 1.

Abstract: A user’s interest in a web page can be estimated by unobtrusively (implicitly) observing his or her behaviour rather than asking for feedback directly (explicitly). Implicit methods are naturally less accurate than explicit methods, but they do not waste a user’s time or effort. Implicit indicators of a user’s interests can also be used to create models that change with a user’s interests over time. Research has shown that a user’s behaviour is related to his/her interest in a web page. We evaluate previously studied implicit indicators and examine the time spent on a page in more detail. For example, we observe whether a user is really looking at the monitor when we measure the time spent on a web page. Our results indicate that the duration is related to a user’s interest of a web page regardless a user’s attention to the web page. 1

this document more interesting than the previous document that was closed within some seconds

Abstract. Search engines return results mainly based on the submitted query; however, the same query could be in different contexts because individual users have different interests. To improve the relevance of search results, we propose re-ranking results based on a learned user profile. In our previous work we introduced a scoring function for re-ranking search results based on a learned User Interest Hierarchy (UIH). Our results indicate that we can improve relevance at lower ranks, but not at the top 5 ranks. In this paper, we improve the scoring function by incorporating new term characteristics, image characteristics, and pivoted length normalization. Our experimental evaluation shows that the proposed approach can improve relevance in each of the top 10 ranks. 1