The task of identifying cover songs has formerly been studied in terms of a prototypical query retrieval framework. However, this framework is not the only one the task allows. In this article, we revise the task of identifying cover songs to include the notion of sets (or groups) of covers. In particular, we study the application of unsupervised clustering and community detection algorithms to detect cover sets. We consider current state-of-the-art algorithms and propose new methods to achieve this goal. Our experiments show that the detection of cover sets is feasible, that it can be performed in a reasonable amount of time, that it does not require extensive parameter tuning, and that it presents certain robustness to inaccurate measurements. Furthermore, we highlight two direct outcomes that naturally arise from the proposed framework revision: increasing the accuracy of query retrieval-based systems and detecting the original song within a set of covers. 1.

As the number of RSS news feeds continue to increase over the Internet, it becomes necessary to minimize the workload of the user who is otherwise required to scan through huge number of news articles to find related articles of interest, which is a tedious and often an impossible task. In order to solve this problem, we present a novel approach, called InFRSS, which consists of a correlation-based phrase matching (CPM) model and a fuzzy compatibility clustering (FCC) model. CPM can detect RSS news articles containing phrases that are the same as well as semantically alike, and dictate the degrees of similarity of any two articles. FCC identifies and clusters non-redundant, closely related RSS news articles based on their degrees of similarity and a fuzzy compatibility relation. Experimental results show that (i) our CPM model on matching bigrams and trigrams in RSS news articles outperforms other phrase/keyword-matching approaches and (ii) our FCC model generates high quality clusters and outperforms other well-known clustering techniques.

Abstract. Tens of thousands of news articles are posted on-line each day, covering topics from politics to science to current events. In order to better cope with this overwhelming volume of information, RSS (news) feeds are used to categorize newly posted articles. Nonetheless, most RSS users must filter through many articles within the same or different RSS feeds in order to locate articles pertaining to their particular interests. Due to the large number of news articles in individual RSS feeds, there is a need for further organizing articles to aid users in locating non-redundant, informative, and related articles of interest quickly. In this paper, we present a novel approach which uses the word-correlation factors in a fuzzy set information retrieval model to (i) filter out redundant news articles from RSS feeds, (ii) shed less-informative articles from the non-redundant ones, and (iii) cluster the remaining informative articles according to the fuzzy equivalence classes generated on the news articles. Our clustering approach requires little overhead or computational costs, and experimental results have shown that it outperforms other existing well-known clustering approaches. 1

Tens of thousands of news articles are posted on-line each day, covering topics from politics to science to current events. To better cope with this overwhelming volume of information, RSS (news) feeds are used to categorize newly posted articles. Nonetheless, most RSS users must filter through many articles within the same or different RSS feeds to locate articles pertaining to their particular interests. Due to the large number of news articles in individual RSS feeds, there is a need for further organizing articles to aid users in locating non-redundant, informative, and related articles of interest quickly. In this paper, we present a novel approach which uses the word-correlation factors in a fuzzy set information retrieval model to (i) filter out redundant news articles from RSS feeds, (ii) shed less-informative articles from the non-redundant ones, and (iii) cluster the remaining informative articles according to the fuzzy equivalence classes on the news articles. Our clustering approach requires little overhead or computational costs, and experimental results have shown that it outperforms other existing, wellknown clustering approaches. The clustering approach as proposed in this paper applies only to RSS news articles; however, it can be extended to other application domains.

The contents of this report are the sole responsibility of the authors. O conteúdo do presente relatório é de única responsabilidade dos autores. A new Hybrid Clustering Approach for Image Retrieval ∗

Dissertação apresentada como requisito parcial para obtenção do grau de Mestre pelo Curso de Mestrado em Ciência da Computação do Centro de Informática

Hierarchical taxonomies provide a multi-level view of large document collections, allowing users to rapidly drill down to finegrained distinctions in topics of interest. We show that automatically induced taxonomies can be made more robust by combining text with relational links. The underlying mechanism is a Bayesian generative model in which a latent hierarchical structure explains the observed data — thus, finding hierarchical groups of documents with similar word distributions and dense network connections. As a nonparametric Bayesian model, our approach does not require pre-specification of the branching factor at each non-terminal, but finds the appropriate level of detail directly from the data. Unlike many prior latent space models of network structure, the complexity of our approach does not grow quadratically in the number of documents, enabling application to networks with more than ten thousand nodes. Experimental results on hypertext and citation network corpora demonstrate the advantages of our hierarchical, multimodal approach.

pavla.drazdilova, jan.martinovic,

Web clustering engines organize search results by topic, thus offering a complementary view to the flat-ranked list returned by conventional search engines. In this survey, we discuss the issues that must be addressed in the development of a Web clustering engine, including acquisition and preprocessing of search results, their clustering and visualization. Search results clustering, the core of the system, has specific requirements that cannot be addressed by classical clustering algorithms. We emphasize the role played by the quality of the cluster labels as opposed to optimizing only the clustering structure. We highlight the main characteristics of a number of existing Web clustering engines and also discuss how to evaluate their retrieval performance. Some directions for future research are finally presented.

In this paper we describe a method for recommending news on a news portal based on our novel representation by a similarity tree. Our method for recommending articles is based on their content. The recommendation employs a hierarchical incremental clustering which is used to discover additional information for effective recommending. The important and novel part of our method is an approach to discovering the interests of individual readers using tree structure created according to similarity of articles. We concentrate on enabling the recommendations in any time, i.e. we discover user’s interests real-time. Our method discovers specific interests of the reader using information gained from monitoring his activities in the news portal. We describe the mechanisms for recommending up-to-date and relevant articles. It is based on known solutions, but incorporates unique representation of user interests by binary tree. Moreover, our aim was to provide recommendations in real-time. Recommendations are thus generated depending on the actual reader’s interest. We also present an evaluation of recommendations in the experiment where we use accounts of real readers and their history of reading. 1