Cai-Nicolas Ziegler

Improving recommendation lists through topic diversification

In this work we present topic diversification, a novel method designed to balance and diversify personalized recommendation lists in order to reflect the users complete spectrum of interests. Though being detrimental to average accuracy, we show that our method improves user satisfaction with recommendation lists, in particular for lists generated using the common item-based collaborative filtering algorithm.Our work builds upon prior research on recommender systems, looking at properties of recommendation lists as entities in their own right rather than specifically focusing on the accuracy of individual recommendations. We introduce the intra-list similarity metric to assess the topical diversity of recommendation lists and the topic diversification approach for decreasing the intra-list similarity. We evaluate our method using book recommendation data, including offline analysis on 361, !, 349 ratings and an online study involving more than 2, !, 100 subjects.

Propagation Models for Trust and Distrust in Social Networks

Semantic Web endeavors have mainly focused on issues pertaining to knowledge representation and ontology design. However, besides understanding information metadata stated by subjects, knowing about their credibility becomes equally crucial. Hence, trust and trust metrics, conceived as computational means to evaluate trust relationships between individuals, come into play. Our major contribution to Semantic Web trust management through this work is twofold. First, we introduce a classification scheme for trust metrics along various axes and discuss advantages and drawbacks of existing approaches for Semantic Web scenarios. Hereby, we devise an advocacy for local group trust metrics, guiding us to the second part which presents Appleseed, our novel proposal for local group trust computation. Compelling in its simplicity, Appleseed borrows many ideas from spreading activation models in psychology and relates their concepts to trust evaluation in an intuitive fashion. Moreover, we provide extensions for the Appleseed nucleus that make our trust metric handle distrust statements.

Investigating interactions of trust and interest similarity

Online communities that allow their users to express their personal preferences, such as the members they trust and the products they appreciate, are becoming increasingly popular. Exploiting these communities as playgrounds for sociological research we present two frameworks for analyzing the correlation between interpersonal trust and interest similarity. We obtain empirical results from applying the two frameworks on two real, operational communities, that suggest there is a strong correlation between both trust and interest similarity. We believe our findings particularly relevant for ongoing research in recommender systems and collaborative filtering, where people are suggested products based on their similarity with other customers, and propose ways in which trust models can be integrated into these systems.

Spreading activation models for trust propagation

Semantic Web endeavors have mainly focused on issues pertaining to knowledge representation and ontology design. However, besides understanding information metadata stated by subjects, knowing about their credibility becomes equally crucial. Hence, trust and trust metrics, conceived as computational means to evaluate trust relationships between individuals, come into play. Our major contributions to semantic Web trust management are twofold. First, we introduce our classification scheme for trust metrics along various axes and discuss advantages and drawbacks of existing approaches for semantic Web scenarios. Hereby, we devise our advocacy for local group trust metrics, guiding us to the second part which presents Appleseed, our novel proposal for local group trust computation. Compelling in its simplicity, Appleseed borrows many ideas from spreading activation models in psychology and relates their concepts to trust evaluation in an intuitive fashion.

Taxonomy-driven computation of product recommendations

Recommender systems have been subject to an enormous rise in popularity and research interest over the last ten years. At the same time, very large taxonomies for product classification are becoming increasingly prominent among e-commerce systems for diverse domains, rendering detailed machine-readable content descriptions feasible. Amazon.com makes use of an entire plethora of hand-crafted taxonomies classifying books, movies, apparel, and various other goods. We exploit such taxonomic background knowledge for the computation of personalized recommendations. Hereby, relationships between super-concepts and sub-concepts constitute an important cornerstone of our novel approach, providing powerful inference opportunities for profile generation based upon the classification of products that customers have chosen. Ample empirical analysis, both offline and online, demonstrates our proposals superiority over common existing approaches when user information is sparse and implicit ratings prevail.

Analyzing Correlation between Trust and User Similarity in Online Communities

Past evidence has shown that generic approaches to recommender systems based upon collaborative filtering tend to poorly scale. Moreover, their fitness for scenarios supposing distributed data storage and decentralized control, like the Semantic Web, becomes largely limited for various reasons. We believe that computational trust models bear several favorable properties for social filtering, opening new opportunities by either replacing or supplementing current techniques. However, in order to provide meaningful results for recommender system applications, we expect notions of trust to clearly reflect user similarity. In this work, we therefore provide empirical results obtained from one real, operational community and verify latter hypothesis for the domain of book recommendations.

Semantic web recommender systems

Research on recommender systems has primarily addressed centralized scenarios and largely ignored open, decentralized systems where remote information distribution prevails The absence of superordinate authorities having full access and control introduces some serious issues requiring novel approaches and methods Hence, our primary objective targets the successful deployment and integration of recommender system facilities for Semantic Web applications, making use of novel technologies and concepts and incorporating them into one coherent framework.

Automatic computation of semantic proximity using taxonomic knowledge

Taxonomic measures of semantic proximity allow us to compute the relatedness of two concepts. These metrics are versatile instruments required for diverse applications, e.g., the Semantic Web, linguistics, and also text mining. However, most approaches are only geared towards hand-crafted taxonomic dictionaries such as WordNet, which only feature a limited fraction of real-world concepts. More specific concepts, and particularly instances of concepts, i.e., names of artists, locations, brand names, etc., are not covered.The contributions of this paper are two fold. First, we introduce a framework based on Google and the Open Directory Project (ODP), enabling us to derive the semantic proximity between arbitrary concepts and instances. Second, we introduce a new taxonomy-driven proximity metric tailored for our framework. Studies with human subjects corroborate our hypothesis that our new metric outperforms benchmark semantic proximity metrics and comes close to human judgement.

Content Extraction from News Pages Using Particle Swarm Optimization on Linguistic and Structural Features

Todays Web pages are commonly made up of more than merely one cohesive block of information. For instance, news pages from popular media channels such as Financial Times or Washington Post consist of no more than 30%-50% of textual news, next to advertisements, link lists to related articles, disclaimer information, and so forth. However, for many search-oriented applications such as the detection of relevant pages for an in-focus topic, dissecting the actual textual content from surrounding page clutter is an essential task, so as to maintain appropriate levels of document retrieval accuracy. We present a novel approach that extracts real content from news Web pages in an unsupervised fashion. Our method is based on distilling linguistic and structural features from text blocks in HTML pages, having a particle swarm optimizer (PSO) learn feature thresholds for optimal classification performance. Empirical evaluations and benchmarks show that our approach works very well when applied to several hundreds of news pages from popular media in 5 languages.Todays Web pages are commonly made up of more than merely one cohesive block of information. For instance, news pages from popular media channels such as Financial Times or Washington Post consist of no more than 30%-50% of textual news, next to advertisements, link lists to related articles, disclaimer information, and so forth. However, for many search-oriented applications such as the detection of relevant pages for an in-focus topic, dissecting the actual textual content from surrounding page clutter is an essential task, so as to maintain appropriate levels of document retrieval accuracy. We present a novel approach that extracts real content from news Web pages in an unsupervised fashion. Our method is based on distilling linguistic and structural features from text blocks in HTML pages, having a Particle Swarm Optimizer (PSO) learn feature thresholds for optimal classification performance. Empirical evaluations and benchmarks show that our approach works very well when applied to several hundreds of news pages from popular media in 5 languages.

Towards Automated Reputation and Brand Monitoring on the Web

The ever-increasing growth of the Web as principal provider of news and opinions makes it impossible for individuals to manually spot and analyze all information of particular importance for global large-scale corporations. Hence, automated means, identifying upcoming topics of utter relevance and monitoring the reputation of a brand as well as its competitors, are becoming indispensable. In this paper, we present a platform for analyzing Web data for such purposes, adopting different semantic perspectives and providing the market analyst with a flexible suite of instruments. We focus on two of these tools and outline their particular utility for research and exploration