Guillaume Cleuziou

An extended version of the k-means method for overlapping clustering

This paper deals with overlapping clustering, a trade off between crisp and fuzzy clustering. It has been motivated by recent applications in various domains such as information retrieval or biology. We show that the problem of finding a suitable coverage of data by overlapping clusters is not a trivial task. We propose a new objective criterion and the associated algorithm OKM that generalizes the k-means algorithm. Experiments show that overlapping clustering is a good alternative and indicate that OKM outperforms other existing methods.

CoFKM: A Centralized Method for Multiple-View Clustering

This paper deals with clustering for multi-view data, i.e. objects described by several sets of variables or proximity matrices. Many important domains or applications such as Information Retrieval, biology, chemistry and marketing are concerned by this problematic. The aim of this data mining research field is to search for clustering patterns that perform a consensus between the patterns from different views. This requires to merge information from each view by performing a fusion process that identifies the agreement between the views and solves the conflicts. Various fusion strategies can be applied, occurring either before, after or during the clustering process. We draw our inspiration from the existing algorithms based on a centralized strategy. We propose a fuzzy clustering approach that generalizes the three fusion strategies and outperforms the main existing multi-view clustering algorithm both on synthetic and real datasets.

Query log driven web search results clustering

Different important studies in Web search results clustering have recently shown increasing performances motivated by the use of external resources. Following this trend, we present a new algorithm called Dual C-Means, which provides a theoretical background for clustering in different representation spaces. Its originality relies on the fact that external resources can drive the clustering process as well as the labeling task in a single step. To validate our hypotheses, a series of experiments are conducted over different standard datasets and in particular over a new dataset built from the TREC Web Track 2012 to take into account query logs information. The comprehensive empirical evaluation of the proposed approach demonstrates its significant advantages over traditional clustering and labeling techniques.

Overview of Overlapping Partitional Clustering Methods

Identifying non-disjoint clusters is an important issue in clustering referred to as Overlapping Clustering. While traditional clustering methods ignore the possibility that an observation can be assigned to several groups and lead to k exhaustive and exclusive clusters representing the data, Overlapping Clustering methods offer a richer model for fitting existing structures in several applications requiring a non-disjoint partitioning. In fact, the issue of overlapping clustering has been studied since the last four decades leading to several methods in the literature adopting many usual approaches such as hierarchical, generative, graphical and k-means based approach. We review in this paper the fundamental concepts of overlapping clustering while we survey the widely known overlapping partitional clustering algorithms and the existing techniques to evaluate the quality of non-disjoint partitioning. Furthermore, a comparative theoretical and experimental study of used techniques to model overlaps is given over different multi-labeled benchmarks.

Biology Based Alignments of Paraphrases for Sentence Compression

In this paper, we present a study for extracting and aligning paraphrases in the context of Sentence Compression. First, we justify the application of a new measure for the automatic extraction of paraphrase corpora. Second, we discuss the work done by (Barzilay & Lee, 2003) who use clustering of paraphrases to induce rewriting rules. We will see, through classical visualization methodologies (Kruskal & Wish, 1977) and exhaustive experiments, that clustering may not be the best approach for automatic pattern identification. Finally, we will provide some results of different biology based methodologies for pairwise paraphrase alignment.

Two Variants of the OKM for Overlapping Clustering

This paper deals with overlapping clustering and presents two extensions of the approach OKM denoted as OKMED andWOKM. OKMED generalizes the well known k-medoid method to overlapping clustering and help in organizing data with any proximity matrix as input. WOKM (Weighted-OKM) proposes a model with local weighting of the clusters; this variant is suitable for overlapping clustering since a single data can matches with multiple classes according to different features. On text clustering, we show that OKMED has a behavior similar to OKM but offers to use metrics other than euclidean distance. Then we observe significant improvement using the weighted extension of OKM.

QASSIT at SemEval-2016 Task 13: On the Integration of Semantic Vectors in Pretopological Spaces for Lexical Taxonomy Acquisition

This paper presents our participation to the SemEval “Task 13: Taxonomy Extraction Evaluation (TExEval-2)” (Bordea et al., 2016). This year, we propose the combination of recent semantic vectors representation into a methodology for semisupervised and auto-supervised acquisition of lexical taxonomies from raw texts. In our proposal, first similarities between concepts are calculated using semantic vectors, then a pretopological space is defined from which a preliminary structure is constructed. Finally, a genetic algorithm is used to optimize two different functions, the quality of the added relationships in the taxonomy and the quality of the structure. Experiments show that our proposal has a competitive performance when compared with the other participants achieving the second position in the general rank.

Informative Polythetic Hierarchical Ephemeral Clustering

Ephemeral clustering has been studied for more than a decade, although with low user acceptance. According to us, this situation is mainly due to (1) an excessive number of generated clusters, which makes browsing difficult and (2) low quality labeling, which introduces imprecision within the search process. In this paper, our motivation is twofold. First, we propose to reduce the number of clusters of Web page results, but keeping all different query meanings. For that purpose, we propose a new polythetic methodology based on an informative similarity measure, the InfoSimba, and a new hierarchical clustering algorithm, the HISGK-means. Second, a theoretical background is proposed to define meaningful cluster labels embedded in the definition of the HISGK-means algorithm, which may elect as best label, words outside the given cluster. To confirm our intuitions, we propose a new evaluation framework, which shows that we are able to extract most of the important query meanings but generating much less clusters than state-of-the-art systems.

Genre and domain processing in an information retrieval perspective

The massive amount of textual data on the Web raises numerous classification problems. Although the notion of domain is widely acknowledged in the IR field, the applicative concept of genre could solve its weaknesses by taking into account the linguistic properties and the document structures of the texts. Two clustering methods are proposed here to illustrate the complementarity of the notions to characterize a closed scientific article corpus. The results are planned to be used in a Web-based application.

Disjunctive Learning with a Soft-Clustering Method

In the case of concept learning from positive and negative examples, it is rarely possible to find a unique discriminating conjunctive rule; in most cases, a disjunctive description is needed. This problem, known as disjunctive learning, is mainly solved by greedy methods, iteratively adding rules until all positive examples are covered. Each rule is determined by discriminating properties, where the discriminating power is computed from the learning set. Each rule defines a subconcept of concept to be learned with these methods. The final set of sub-concepts is then highly dependent from both the learning set and the learning method.