Collaborative tagging applications allow Internet users to annotate resources with personalized tags. The complex network created by many annotations, often called a folksonomy, permits the freedom explore tags, or even other user's profiles unbound from rigid predefined conceptual hierarchy. However, afforded comes at cost: an uncontrolled vocabulary can result in tag redundancy and ambiguity hindering navigation. Data mining techniques, such as clustering, provide means remedy these...

Publicly accessible adaptive systems such as collaborative recommender present a security problem. Attackers, who cannot be readily distinguished from ordinary users, may inject biased profiles in an attempt to force system “adapt” manner advantageous them. Such attacks lead degradation of user trust the objectivity and accuracy system. Recent research has begun examine vulnerabilities robustness different recommendation techniques face “profile injection” attacks. In this article, we...

Many recommendation algorithms suffer from popularity bias in their output: popular items are recommended frequently and less ones rarely, if at all. However, popular, long-tail precisely those that often desirable recommendations. In this paper, we introduce a flexible regularization-based framework to enhance the coverage of lists learning-to-rank algorithm. We show regularization provides tunable mechanism for controlling trade-off between accuracy coverage. Moreover, experimental results...

Every user has a distinct background and specific goal when searching for information on the Web. The of Web search personalization is to tailor results particular based that user's interests preferences. Effective access involves two important challenges: accurately identifying context organizing in such way matches context. We present an approach personalized building models as ontological profiles by assigning implicitly derived interest scores existing concepts domain ontology. A...

Contextual factors can greatly influence the users' preferences in listening to music. Although it is hard capture these directly, possible see their effects on sequence of songs liked by user his/her current interaction with system. In this paper, we present a context-aware music recommender system which infers contextual information based most recent user. Our approach mines top frequent tags for from social tagging Web sites and uses topic modeling determine set latent topics each song,...

Recommendation algorithms are known to suffer from popularity bias; a few popular items recommended frequently while the majority of other ignored. These recommendations then consumed by users, their reaction will be logged and added system: what is generally as feedback loop. In this paper, we propose method for simulating users interaction with recommenders in an offline setting study impact loop on bias amplification several recommendation algorithms. We show how leads problems such...

This technical report describes FAQ Finder, a natural language question answering system that uses files of frequently asked questions as its knowledge base. Unlike AI question-answering systems focus on the generation new answers, Finder retrieves existing ones found in frequently-asked files. information retrieval approaches rely purely lexical metric similarity between query and document, semantic base (WordNet) to improve ability match answer. We describe design current implementation...

While the explosion of online information has introduced new opportunities for finding and using electronic data, it also underscored problem isolating useful making sense large, multidimensional spaces. In response to this problem, we have developed an approach building data tour guides, called FindMe systems. These programs know enough about space help users navigate through it, sure they not only come away with but insights into structure itself. these systems, combined idea...

Recommender systems support users in identifying products and services e-commerce other information-rich environments. Recommendation problems have a long history as successful AI application area, with substantial interest beginning the mid-1990s, increasing subsequent rise of e-commerce. research focused on recommending only simple such movies or books; constraint-based recommendation now receives attention due to capability complex services. In this paper, we first introduce taxonomy...

Collaborative recommender systems are highly vulnerable to attack. Attackers can use automated means inject a large number of biased profiles into such system, resulting in recommendations that favor or disfavor given items. Since collaborative must be open user input, it is difficult design system cannot so attacked. Researchers studying robust recommendation have therefore begun identify types attacks and study mechanisms for recognizing defeating them. In this paper, we propose different...

Recommender systems are tools for interacting with large and complex information spaces. They provide a personalized view of such spaces, prioritizing items likely to be interest the user. The field, christened in 1995, has grown enormously variety problems addressed techniques employed, as well its practical applications. research incorporated wide artificial intelligence including machine learning, data mining, user modeling, case‐based reasoning, constraint satisfaction, among others....

Recently there has been a growing interest in fairness-aware recommender systems including fairness providing consistent performance across different users or groups of users. A system could be considered unfair if the recommendations do not fairly represent tastes certain group while other receive that are with their preferences. In this paper, we use metric called miscalibration for measuring how recommendation algorithm is responsive to users' true preferences and consider various...

Recommender system has been demonstrated as one of the most useful tools to assist users' decision makings. Several recommendation algorithms have developed and implemented by both commercial open-source libraries. Context-aware recommender (CARS) emerged a novel research direction during past decade many contextual proposed. Unfortunately, no engines start embed those in their kits, due special characteristics data format processing methods domain CARS. This paper introduces an Java-based...

Recommendation, information retrieval, and other access systems pose unique challenges for investigating applying the fairness non-discrimination concepts that have been developed studying machine learning systems.While fair shares many commonalities with classi cation, there are important di erences: multistakeholder nature of applications, rank-based problem se ing, centrality personalization in cases, role user response all complicate identifying precisely what types operationalizations...

Collaborative-filtering recommender systems are an electronic extension of everyday social recommendation behavior: people share opinions and decide whether or not to act on the basis what they hear. Collaborative filtering lets you scale such interactions groups thousands even millions. Publicly accessible user-adaptive as collaborative introduce security issues that must be solved if users perceive these objective, unbiased, accurate.