arXiv subjects

Vincent Labatut

Publications and source records attributed to Vincent Labatut.

70 records · Page 4Linked to original sources

arXiv2013

Rôle communautaire des capitalistes sociaux dans Twitter

Les capitalistes sociaux sont des utilisateurs de médias sociaux tels que Twitter, appliquant diverses techniques pour obtenir un maximum de visibilité. Ils peuvent être néfastes à l'équilibre du service, dans la mesure où leurs comptes, en gagnant en importance sans réelle raison de contenu, rendent difficile l'accès à un contenu pertinent. Dans ce travail, nous nous intéressons à leur caractérisation d'un point de vue purement topologique, i.e. sans considérer la nature des contenus partagés. Nous utilisons pour cela la notion de rôle communautaire, qui est basée sur la structure de communautés du réseau étudié. Nous apportons des modifications à des mesures précédemment définies à cet effet, et proposons une méthode objective de détection des rôles. Nous appliquons ensuite notre méthode à l'analyse d'un réseau représentant Twitter. Nos résultats montrent que les rôles que nous identifions via nos mesures se révèlent particulièrement cohérents par rapport aux capitalistes sociaux du réseau Twitter, dont le comportement est clairement identifié---Social capitalists are social media users taking advantage of various methods to maximize their visibility. This results in artificially important accounts, in the sense this importance is not backed by any real content. The risk is then to see those accounts hiding relevant contents and therefore preventing other users to access them. In this work, we want to characterize social capitalists from a purely topological perspective, i.e. without considering the nature of the shared contents. For this purpose, we use the notion of community role, based on the community structure of the studied network. We modify some measures previously designed for this matter, and propose an objective method to determine roles. We then apply this method to the analysis of a Twitter network. Our results show the roles identified through our measures are particularly consistent with Twitter's social capitalists, whose behavior was clearly identified.

cs.CY↗

arXiv2013

A Comparison of Named Entity Recognition Tools Applied to Biographical Texts

Named entity recognition (NER) is a popular domain of natural language processing. For this reason, many tools exist to perform this task. Amongst other points, they differ in the processing method they rely upon, the entity types they can detect, the nature of the text they can handle, and their input/output formats. This makes it difficult for a user to select an appropriate NER tool for a specific situation. In this article, we try to answer this question in the context of biographic texts. For this matter, we first constitute a new corpus by annotating Wikipedia articles. We then select publicly available, well known and free for research NER tools for comparison: Stanford NER, Illinois NET, OpenCalais NER WS and Alias-i LingPipe. We apply them to our corpus, assess their performances and compare them. When considering overall performances, a clear hierarchy emerges: Stanford has the best results, followed by LingPipe, Illionois and OpenCalais. However, a more detailed evaluation performed relatively to entity types and article categories highlights the fact their performances are diversely influenced by those factors. This complementarity opens an interesting perspective regarding the combination of these individual tools in order to improve performance.

cs.IR↗

arXiv2013

Towards realistic artificial benchmark for community detection algorithms evaluation

Assessing the partitioning performance of community detection algorithms is one of the most important issues in complex network analysis. Artificially generated networks are often used as benchmarks for this purpose. However, previous studies showed their level of realism have a significant effect on the algorithms performance. In this study, we adopt a thorough experimental approach to tackle this problem and investigate this effect. To assess the level of realism, we use consensual network topological properties. Based on the LFR method, the most realistic generative method to date, we propose two alternative random models to replace the Configuration Model originally used in this algorithm, in order to increase its realism. Experimental results show both modifications allow generating collections of community-structured artificial networks whose topological properties are closer to those encountered in real-world networks. Moreover, the results obtained with eleven popular community identification algorithms on these benchmarks show their performance decrease on more realistic networks.

cs.SI↗

arXiv2013

On Flexible Web Services Composition Networks

The semantic Web service community develops efforts to bring semantics to Web service descriptions and allow automatic discovery and composition. However, there is no widespread adoption of such descriptions yet, because semantically defining Web services is highly complicated and costly. As a result, production Web services still rely on syntactic descriptions, key-word based discovery and predefined compositions. Hence, more advanced research on syntactic Web services is still ongoing. In this work we build syntactic composition Web services networks with three well known similarity metrics, namely Levenshtein, Jaro and Jaro-Winkler. We perform a comparative study on the metrics performance by studying the topological properties of networks built from a test collection of real-world descriptions. It appears Jaro-Winkler finds more appropriate similarities and can be used at higher thresholds. For lower thresholds, the Jaro metric would be preferable because it detect less irrelevant relationships.

cs.SE↗

arXiv2013

Topological Properties of Web Services Similarity Networks

The number of publicly available Web services (WS) is continuously growing. To perform efficient WS discovery, it is desirable to organize the WS space. Works in this direction propose to group WS according to certain shared properties. Such groups commonly called communities are based either on similarity or on interaction between WS. In this paper we focus on the former, and propose a new network-based approach to extract communities from a WS collection. This process is three-stepped: first we define several similarity functions able to compare WS operations, second we use them to build so-called similarity networks, and third we identify communities under the form of specific structures in these networks. We apply our method on a collection of real-world WS and comment the resulting communities. Finally, we additionally provide an analysis and an interpretation of our similarity networks with a complex networks perspective.

cs.IR↗

arXiv2013

Web Services Dependency Networks Analysis

Along with a continuously growing number of publicly available Web services (WS), we are witnessing a rapid development in semantic-related web technologies, which lead to the apparition of semantically described WS. In this work, we perform a comparative analysis of the syntactic and semantic approaches used to describe WS, from a complex network perspective. First, we extract syntactic and semantic WS dependency networks from a collection of publicly available WS descriptions. Then, we take advantage of tools from the complex network field to analyze them and determine their topological properties. We show WS dependency networks exhibit some of the typical characteristics observed in real-world networks, such as small world and scale free properties, as well as community structure. By comparing syntactic and semantic networks through their topological properties, we show the introduction of semantics in WS description allows modeling more accurately the dependencies between parameters, which in turn could lead to improved composition mining methods.

cs.IR↗

arXiv2013

Benefits of Semantics on Web Service Composition from a Complex Network Perspective

The number of publicly available Web services (WS) is continuously growing, and in parallel, we are witnessing a rapid development in semantic-related web technologies. The intersection of the semantic web and WS allows the development of semantic WS. In this work, we adopt a complex network perspective to perform a comparative analysis of the syntactic and semantic approaches used to describe WS. From a collection of publicly available WS descriptions, we extract syntactic and semantic WS interaction networks. We take advantage of tools from the complex network field to analyze them and determine their properties. We show that WS interaction networks exhibit some of the typical characteristics observed in real-world networks, such as short average distance between nodes and community structure. By comparing syntactic and semantic networks through their properties, we show the introduction of semantics in WS descriptions should improve the composition process.

cs.SI↗

arXiv2013

MATAWS: A Multimodal Approach for Automatic WS Semantic Annotation

Many recent works aim at developing methods and tools for the processing of semantic Web services. In order to be properly tested, these tools must be applied to an appropriate benchmark, taking the form of a collection of semantic WS descriptions. However, all of the existing publicly available collections are limited by their size or their realism (use of randomly generated or resampled descriptions). Larger and realistic syntactic (WSDL) collections exist, but their semantic annotation requires a certain level of automation, due to the number of operations to be processed. In this article, we propose a fully automatic method to semantically annotate such large WS collections. Our approach is multimodal, in the sense it takes advantage of the latent semantics present not only in the parameter names, but also in the type names and structures. Concept-to-word association is performed by using Sigma, a mapping of WordNet to the SUMO ontology. After having described in details our annotation method, we apply it to the larger collection of real-world syntactic WS descriptions we could find, and assess its efficiency.

cs.SE↗

arXiv2012

Une nouvelle mesure pour l'évaluation des méthodes de détection de communautés

Community detection can be considered as a variant of cluster analysis applied to complex networks. For this reason, all existing studies have been using tools derived from this field when evaluating community detection algorithms. However, those are not completely relevant in the context of network analysis, because they ignore a part of the available information, and can therefore lead to incorrect interpretations. In this article, we illustrate this limitation, and propose a solution by modifying an existing measure. We then apply it to realistic community-structured networks, in order to perform a first evaluation.---La détection de communautés dans un réseau complexe est une tâche que l'on peut rapprocher de la classification non-supervisée réalisée en fouille de données classique. Pour cette raison, l'évaluation des algorithmes accomplissant ce type de traitement s'est faite jusqu'ici exclusivement au moyen de mesures comparables à celles utilisées en fouille de données. Cependant, dans le cas de l'analyse de réseau, celles-ci n'exploitent pas toute l'information disponible et sont susceptibles de fournir des résultats biaisés. Dans cet article, nous illustrons cette limitation et proposons une solution en modifiant une mesure existante. Nous l'appliquons ensuite à des données réalistes afin d'en effectuer une première évaluation expérimentale.

math.ST↗

arXiv2012

Topological measures for the analysis of wireless sensor networks

Concepts such as energy dependence, random deployment, dynamic topological update, self-organization, varying large number of nodes are among many factors that make WSNs a type of complex system. However, when analyzing WSNs properties using complex network tools, classical topological measures must be considered with care as they might not be applicable in their original form. In this work, we focus on the topological measures frequently used in the related field of Internet topological analysis. We illustrate their applicability to the WSNs domain through simulation experiments. In the cases when the classic metrics turn out to be incompatible, we propose some alternative measures and discuss them based on the WSNs characteristics.

cs.NI↗

arXiv2012

Accuracy Measures for the Comparison of Classifiers

The selection of the best classification algorithm for a given dataset is a very widespread problem. It is also a complex one, in the sense it requires to make several important methodological choices. Among them, in this work we focus on the measure used to assess the classification performance and rank the algorithms. We present the most popular measures and discuss their properties. Despite the numerous measures proposed over the years, many of them turn out to be equivalent in this specific case, to have interpretation problems, or to be unsuitable for our purpose. Consequently, classic overall success rate or marginal rates should be preferred for this specific task.

cs.LG↗

arXiv2012

Qualitative Comparison of Community Detection Algorithms

Community detection is a very active field in complex networks analysis, consisting in identifying groups of nodes more densely interconnected relatively to the rest of the network. The existing algorithms are usually tested and compared on real-world and artificial networks, their performance being assessed through some partition similarity measure. However, artificial networks realism can be questioned, and the appropriateness of those measures is not obvious. In this study, we take advantage of recent advances concerning the characterization of community structures to tackle these questions. We first generate networks thanks to the most realistic model available to date. Their analysis reveals they display only some of the properties observed in real-world community structures. We then apply five community detection algorithms on these networks and find out the performance assessed quantitatively does not necessarily agree with a qualitative analysis of the identified communities. It therefore seems both approaches should be applied to perform a relevant comparison of the algorithms.

cs.SI↗

arXiv2012

An Empirical Study of the Relation Between Community Structure and Transitivity

One of the most prominent properties in real-world networks is the presence of a community structure, i.e. dense and loosely interconnected groups of nodes called communities. In an attempt to better understand this concept, we study the relationship between the strength of the community structure and the network transitivity (or clustering coefficient). Although intuitively appealing, this analysis was not performed before. We adopt an approach based on random models to empirically study how one property varies depending on the other. It turns out the transitivity increases with the community structure strength, and is also affected by the distribution of the community sizes. Furthermore, increasing the transitivity also results in a stronger community structure. More surprisingly, if a very weak community structure causes almost zero transitivity, the opposite is not true and a network with a close to zero transitivity can still have a clearly defined community structure. Further analytical work is necessary to characterize the exact nature of the identified relationship.

cs.SI↗

arXiv2012

Comparative Evaluation of Community Detection Algorithms: A Topological Approach

Community detection is one of the most active fields in complex networks analysis, due to its potential value in practical applications. Many works inspired by different paradigms are devoted to the development of algorithmic solutions allowing to reveal the network structure in such cohesive subgroups. Comparative studies reported in the literature usually rely on a performance measure considering the community structure as a partition (Rand Index, Normalized Mutual information, etc.). However, this type of comparison neglects the topological properties of the communities. In this article, we present a comprehensive comparative study of a representative set of community detection methods, in which we adopt both types of evaluation. Community-oriented topological measures are used to qualify the communities and evaluate their deviation from the reference structure. In order to mimic real-world systems, we use artificially generated realistic networks. It turns out there is no equivalence between both approaches: a high performance does not necessarily correspond to correct topological properties, and vice-versa. They can therefore be considered as complementary, and we recommend applying both of them in order to perform a complete and accurate assessment.

cs.SI↗

arXiv2011

Evaluation of Performance Measures for Classifiers Comparison

The selection of the best classification algorithm for a given dataset is a very widespread problem, occuring each time one has to choose a classifier to solve a real-world problem. It is also a complex task with many important methodological decisions to make. Among those, one of the most crucial is the choice of an appropriate measure in order to properly assess the classification performance and rank the algorithms. In this article, we focus on this specific task. We present the most popular measures and compare their behavior through discrimination plots. We then discuss their properties from a more theoretical perspective. It turns out several of them are equivalent for classifiers comparison purposes. Futhermore. they can also lead to interpretation problems. Among the numerous measures proposed over the years, it appears that the classical overall success rate and marginal rates are the more suitable for classifier comparison task.

cs.LG↗

arXiv2011

On Accuracy of Community Structure Discovery Algorithms

Community structure discovery in complex networks is a quite challenging problem spanning many applications in various disciplines such as biology, social network and physics. Emerging from various approaches numerous algorithms have been proposed to tackle this problem. Nevertheless little attention has been devoted to compare their efficiency on realistic simulated data. To better understand their relative performances, we evaluate systematically eleven algorithms covering the main approaches. The Normalized Mutual Information (NMI) measure is used to assess the quality of the discovered community structure from controlled artificial networks with realistic topological properties. Results show that along with the network size, the average proportion of intra-community to inter-community links is the most influential parameter on performances. Overall, "Infomap" is the leading algorithm, followed by "Walktrap", "SpinGlass" and "Louvain" which also achieve good consistency.

cs.SI↗