This paper examines the reliability of implicit feedback generated from clickthrough data in WWW search. Analyzing the users ’ decision process using eyetracking and comparing implicit feedback against manual relevance judgments, we conclude that clicks are informative but biased. While this makes the interpretation of clicks as absolute relevance judgments difficult, we show that relative preferences derived from clicks are reasonably accurate on average. Categories and Subject Descriptors

This paper examines the reliability of implicit feedback generated from clickthrough data and query reformulations in WWW search. Analyzing the users ’ decision process using eyetracking and comparing implicit feedback against manual relevance judgments, we conclude that clicks are informative but biased. While this makes the interpretation of clicks as absolute relevance judgments difficult, we show that relative preferences derived from clicks are reasonably accurate on average. We find that such relative preferences are accurate not only between results from an individual query, but across multiple sets of results within chains of query reformulations.

As the competition of Web search market increases, there is a high demand for personalized Web search to conduct retrieval incorporating Web users' information needs. This paper focuses on utilizing clickthrough data to improve Web search. Since millions of searches are conducted everyday, a search engine accumulates a large volume of clickthrough data, which records who submits queries and which pages he/she clicks on. The clickthrough data is highly sparse and contains di#erent types of objects (user, query and Web page), and the relationships among these objects are also very complicated. By performing analysis on these data, we attempt to discover Web users' interests and the patterns that users locate information. In this paper, a novel approach CubeSVD is proposed to improve Web search. The clickthrough data is represented by a 3-order tensor, on which we perform 3-mode analysis using the higher-order singular value decomposition technique to automatically capture the latent factors that govern the relations among these multi-type objects: users, queries and Web pages. A tensor reconstructed based on the CubeSVD analysis reflects both the observed interactions among these objects and the implicit associations among them. Therefore, Web search activities can be carried out based on CubeSVD analysis. Experimental evaluations using a real-world data set collected from an MSN search engine show that CubeSVD achieves encouraging search results in comparison with some standard methods.

Abstract. In this paper we make a case for the role of user context information in supporting task performance online, examine previous attempts at representing and making use of user context factors, and highlight the limitations of existing tools and services. We then suggest how the emergent Semantic Web might be able to better facilitate the capture of knowledge regarding user context, and provide the means for its reuse in supporting the performance of tasks online. 1

Users ’ past search behaviour provides a rich context that an information retrieval system can use to tailor its search results to suit an individual’s or a community’s information needs. In this paper, we present an investigation of the variability in search behaviours for the same queries in a close-knit community. By examining web proxy cache logs over a period of nine months, we extracted a set of 135 queries that had been issued by at least ten users. Our analysis indicates that, overall, users clicked on highly ranked and relevant pages, but they tend to click on different sets of pages. Examination of the query reformulation history revealed that users often have different search intents behind the same query. We identify three major causes for the community’s interaction behaviour differences: the variance of task, the different intents expressed with the query, and the snippet and characteristics of retrieved documents. Based on our observations, we identify opportunities to improve the design of different search and delivery tools to better support community and individual search experience.

Finding out about a topic online can be time consuming. It involves visiting multiple news sites, encyclopedia entries, video repositories and other resources while discarding irrelevant information. MakeMyPage aims to speed this process by combining automatic aggregation of information with social media to build web pages with images, videos and links to important information about a topic. MakeMyPage uses automatic aggregation to provide the initial content of the web pages. This content is organized by type: blogs, news, web links, images, video and a main article. MakeMyPage creates a web page by selecting a few items from each category, plus links to more resources within it. Users can vote on the links and media they like best for a given topic and, based on these votes, the system promotes them to and within the main web page. MakeMyPage can be thought of as a collection of wiki pages where people enhance automatically generated content not by editing the text in it, but by voting and suggesting new links. The system’s focus is on the organization of content that is genuinely useful and on point. MakeMyPage continuously tracks popular search queries and maintains a database of web pages about these topics.

Abstract. With the exponential growth of the available information on the World Wide Web, a traditional search engine, even if based on sophisticated document indexing algorithms, has difficulty meeting efficiency and effectiveness performance demanded by users searching for relevant information. Users surfing the Web in search of resources to satisfy their information needs have less and less time and patience to formulate queries, wait for the results and sift through them. Consequently, it is vital in many applications- for example in an e-commerce Web site or in a scientific one- for the search system to find the right information very quickly. Personalized Web environments that build models of short-term and long-term user needs based on user actions, browsed documents or past queries are playing an increasingly crucial role: they form a winning combination, able to satisfy the user better than unpersonalized search engines based on traditional Information Retrieval (IR) techniques. Several important user personalization approaches and techniques developed for the Web search domain are illustrated in this chapter, along with examples of real systems currently being used on the

This report consists of three parts. Part I reviews how users’ online tasks have been conceptualised in previous literature, and how researchers have defined and used context in support of user tasks. Novel conceptualisations of user tasks online and user context factors are then presented and contrasted with earlier work, before a discussion of how these context factors have been supported in previous applications. The modelling of social context is then considered in greater detail, with particular focus on aspects such as the nature of social relationships and trust between individuals. Research gaps identified through this review of the literature are summarised to conclude this section. Part II addresses specific outputs of the research to date. In particular, the conceptualisations of user tasks online and user contexts are discussed in more detail, including coverage of the assumptions they are based upon and the background to their development. Specific technical work carried out is also described, including the planning of a social context application, analysis of tools and technologies that may be utilised, and development of relevant technical skills. Drawing on the gaps identified in Part I of the report, Part III introduces the questions that will be addressed by the research. After justifying the research questions, the methods that will be used are outlined and discussed, including overall plans for how the research will be carried out.

Abstract. We propose a new approach for enhancing collaborative information retrieval by means of incorporating positional data for a location-aware personalized retrieval process. In our framework, the collaboration between users will be established by building communities based on matching user attributes in a uniform user model. This allows for incorporating automated intra-community collaboration into the retrieval process. In addition, continuously changing location-based similarity measures are employed with respect to queries posed using mobile devices in order to enhance the quality of the community dependent answer rankings. The consideration of continuously arriving user positions, however, leads to a high-frequency stream of data. For the efficient processing and analysis of this stream, incremental data stream processing techniques are employed. Our interdisciplinary approach incorporates both techniques from information retrieval and data stream processing to achieve an extended retrieval process in a collaborative environment. 1

Vast information sources like the Internet are very hard to browse or search, and using existing tools might be sometimes frustrating for many people. A number of techniques has been proposed to deliver user-oriented solutions. In this paper we present how query refinement, collaborative filtering and faceted browsing can be enriched with social networking and semantic web technologies, and brought together in one search and browsing information system. We claim that by introducing social networking and semantically enhanced search and browsing techniques the quality and user-friendliness of the information system is increased compared to classical, keyword-based query processing.