moi I am a permanent researcher at CNRS and member of the laboratory ISP at ENS Cachan since Octobre 2015. My work mainly concerns the analysis and modeling of large networks encountered in practice, such as the Internet, the web, social networks or legal networks. These networks are typically represented by graphs and a set of techniques stemming from graph theory has already proven to be very useful to characterize and analyze their structure and to model their behavior by highlighting specific properties. My work follows this approach by proposing new theoretical tools in order to identify non-trivial properties of these networks and to define new models able to capture these properties.
ACM International Conference on the Theory of Information Retrieval (ICTIR)
Document coherence describes how much sense text makes in terms of its logical organisation and d... more Document coherence describes how much sense text makes in terms of its logical organisation and discourse flow. Even though coherence is a relatively difficult notion to quantify precisely, it can be approximated automatically. This type of coherence modelling is not only interesting in itself, but also useful for a number of other text processing tasks, including Information Retrieval (IR), where adjusting the ranking of documents according to both their relevance and their coherence has been shown to increase retrieval effectiveness [34, 37]. The state of the art in unsupervised coherence modelling represents documents as bipartite graphs of sentences and discourse entities, and then projects these bipartite graphs into one–mode undirected graphs. However, one–mode projections may incur significant loss of the information present in the original bipartite structure. To address this we present three novel graph metrics that compute document coherence on the original bipartite graph of sentences and entities. Evaluation on standard settings shows that: (i) one of our coherence metrics beats the state of the art in terms of coherence accuracy; and (ii) all three of our coherence metrics improve retrieval effectiveness because, as closer analysis reveals, they capture aspects of document quality that go undetected by both keyword-based standard ranking and by spam filtering. This work contributes to document coherence metrics that are theoretically principled, parameter-free, and useful to IR.
29th International Conference on Legal Knowledge and Information Systems (JURIX'16)
What is a case decided by the European Court of Human Rights about?
The Courts own case database,... more What is a case decided by the European Court of Human Rights about? The Courts own case database, HUDOC , lists all the articles mentioned in a specific case in their metadata. They also supply a number of keywords, but these keywords for the most part are reduced to repeating phrases from the relevant articles. In order to enhance information retrieval about case content, without relying on manual labor and subjective judgment, we propose in this paper a quantitative method that gives a better indication of case content in terms of which articles a given case is more closely associated with. To do so, we rely on the network structure induced by existing case-to-case and case-to-article citations and propose two computational approaches (referred to as MainIn and MainOuT) which result in assigning one representative article to each case. We validate the approach by selecting a sample of important cases and comparing manual investigation of real content of those cases with the MainIn and MainOut articles. Results show that MainIn in particular is able to infer correctly the real content in most of the cases.
—Do case citations reflect the " real " importance of individual judgments for the legal system c... more —Do case citations reflect the " real " importance of individual judgments for the legal system concerned? This question has long been puzzling empirical legal scholars. Existing research typically studies case citation networks as a whole applying traditional network metrics stemming from graph theory. Those approaches are able to detect globally important cases, but since they do not take time explicitly into account, they cannot provide a comprehensive account of the dynamics behind the network structure and its evolution. In this paper we provide such a description, using two node importance metrics that take time into account to study important cases in the Court of Justice of the European Union over time. We then compare cases deemed as important by the metrics, with a set of 50 cases selected by the Court as the most important (landmark) cases. Our contribution is twofold. First, with regard to network science, we show that structural and time-related properties are complementary, and necessary to obtain a complete and nuanced picture of the citation network. Second, with regard to the case law of the Court, this study provides empirical evidence clarifying the motivation of the Court when selecting the landmark cases, revealing the importance of symbolic and historical cases in the selection. In addition, the temporal analysis sheds new light on the network properties specific to the landmark cases that distinguishes them from the rest of the cases. We validate our results by providing legal interpretations that sustain the highlights provided by the proposed network analysis.
Since the early 2000’s, the Internet topology has been an attractive and important research topic... more Since the early 2000’s, the Internet topology has been an attractive and important research topic, either for developing data collection mechanisms, and for analyzing and modeling the network. Beside traditional aspects of the Internet topology (i.e., IP interface, router, and AS levels), recent researches focused on intermediate promising visions of the topology, namely Point-of-Presence (PoP) and subnets (i.e., a set of devices that are located on the same connection medium and that can communicate directly with each other at the link layer). This paper focuses on network subnet discovery by proposing a new tool called TreeNET. One of the key aspects of TreeNET is that it builds a tree representing the way subnets are located with respect to each other. This tree allows TreeNET to obtain additional information on the network, leading to better analysis of the collected data. In this paper, we demonstrate the potential of TreeNET through the evaluation of its key algorithmic steps and the study of measurements collected from the PlanetLab testbed.
The 18th Asia-Pacific Network Operations and Management Symposium
After being the support of the data and voice
convergence, the Internet has become one of the mai... more After being the support of the data and voice convergence, the Internet has become one of the main video providers such as TV-stream. As an alternative to limited or expensive technologies, P2PTV has turned out to be a promising support for such applications. This infrastructure strongly relies on the overlay composed by the peers that consume and diffuse video contents at the same time. Understanding the dynamical properties of this overlay, and in particular how the users switch from one overlay to another, appears to be a key aspect if one wants to improve the quality of P2PTV.
In this paper, we investigate the question of relying on non- invasive measurement techniques to track the presence of users on several channels of P2PTV. Using two datasets obtained by using network measurement on P2PTV infrastructure, we show that such approach contains sufficient information to track the presence of users on several channels. Besides, exploiting the view provided by sliding time windows, we are able to refine the analysis and track users that switch from one channel to another, leading to the detection of super-peers and providing explanations of the different roles they can play in the infrastructure. In addition, by comparing the results obtained on the two datasets, we show how such analyses can shed some light on the evolution of the infrastructure polic
Many studies have proposed to apply artificial intelligence techniques to
legal networks, whether... more Many studies have proposed to apply artificial intelligence techniques to legal networks, whether it be for highlighting legal reasoning, resolving conflict or extracting information from legal databases. In this context, a new line of research has recently emerged which consists in considering legal decisions as elements of complex networks and conduct a structural analysis of the relations between the decisions. It has proved to be efficient for detecting important decisions in legal rulings. In this paper, we follow this approach and propose to extend structural analyses with temporal properties. We define in particular the notion of relative in-degree, temporal distance and average longevity and use those metrics to rank the legal decisions of the two first trials of the International Criminal Court. The results presented in this paper highlight non trivial temporal properties of those legal networks, such as the presence of decisions with an unexpected high longevity, and show the relevance of the proposed relative in-degree property to detect landmark decisions. We validate the outcomes by confronting the results to the one obtained with the standard in-degree property and provide juridical explanations of the decisions identified as important by our approach
For a long time now, researchers have worked on
defining different metrics able to characterize t... more For a long time now, researchers have worked on defining different metrics able to characterize the importance of nodes in networks. Among them, centrality measures have proved to be pertinent as they relate the position of a node in the structure to its ability to diffuse an information efficiently. The case of dynamic networks, in which nodes and links appear and disappear over time, led the community to propose extensions of those classical measures. Yet, they do not investigate the fact that the network structure evolves and that node importance may evolve accordingly. In the present paper, we propose temporal extensions of notions of centrality, which take into account the paths existing at any given time, in order to study the time evolution of nodes’ importance in dynamic networks. We apply this to two datasets and show that the importance of nodes does indeed vary greatly with time. We also show that in some cases it might be meaningless to try to identify nodes that are consistently important over time, thus strengthening the interest of temporal extensions of centrality measures.
2014 Ieee 22nd International Symposium on Modelling Analysis Simulation of Computer and Telecommunication Systems, Sep 9, 2014
The classical approach for Internet topology measurement consists in distributively collecting as... more The classical approach for Internet topology measurement consists in distributively collecting as much data as possible and merging it into one single piece of topology on which are conducted subsequent analysis. Although this approach may seem reasonable, in most cases network measurements performed in this way suffer from some or all of the following limitations: they give only partial views of the networks under concern, these views may be intrinsically biased, and they contain erroneous data due to the measurement tools. Here we present a new tool, named UDP Ping , that relies on a very different approach for the measurement of the Internet topology. Its basic principle is to measure the interface of a given target directed toward a monitor which sends the measurement probe. We demonstrate how to use it to deploy real world-wide measurements that provide reliable (i.e. bias and error free) knowledge of the Internet topology, namely the degree distribution of routers in the core Internet in our example.
Many real-world networks lend themselves to the use of graphs
for analysing and modelling their s... more Many real-world networks lend themselves to the use of graphs for analysing and modelling their structure. But such a simple representation has proven to miss some important and non trivial properties hidden in the bipartite structure of the networks. Recent papers have shown that overlapping properties seem to be present in bipartite networks and that it could explain better the properties observed in simple graphs. This work intends to investigate this question by studying two proposed metrics to account for overlapping structures in bipartite networks. The study, conducted on four dataset stemming from very different contexts (computer science, juridical science and social science), shows that the most popular metrics, the clustering coefficient, turns out to be less relevant that the recent redundancy coefficient to analyse intricate overlapping properties of real networks.
ABSTRACT The development of wireless devices led the scientific community to focus more and more ... more ABSTRACT The development of wireless devices led the scientific community to focus more and more on systems of interaction composed of moving entities. In this context, different models have been proposed in an attempt to capture properties of the observed dynamics. Among those models, the edge-Markovian evolving graph model is appealing since it enables to highlight temporal dependencies in the evolution of the graphs. This model relies on two parameters accounting respectively for the creation and suppression of links in the graph. Thus it assumes that these two parameters are sufficient to characterise the dynamics during all the evolution of the graph. In this paper, we test this hypothesis by confronting the model to 6 datasets recording real traces of evolving networks. In particular, we study the proportion of created and deleted links over the time. The results show that 5 of the 6 case studies present an heterogeneous distribution of those fractions which contradicts the underlying hypothesis of the model. Besides, in order to understand the importance this might have as regard structural properties of real networks, we also study the impact the Markovian model has on the mean degree over the time. It turns out that even in the suitable case, the model fails to reproduce correctly this property which indicates its inadequacy for even more complex properties of real evolving networks.
ACM International Conference on the Theory of Information Retrieval (ICTIR)
Document coherence describes how much sense text makes in terms of its logical organisation and d... more Document coherence describes how much sense text makes in terms of its logical organisation and discourse flow. Even though coherence is a relatively difficult notion to quantify precisely, it can be approximated automatically. This type of coherence modelling is not only interesting in itself, but also useful for a number of other text processing tasks, including Information Retrieval (IR), where adjusting the ranking of documents according to both their relevance and their coherence has been shown to increase retrieval effectiveness [34, 37]. The state of the art in unsupervised coherence modelling represents documents as bipartite graphs of sentences and discourse entities, and then projects these bipartite graphs into one–mode undirected graphs. However, one–mode projections may incur significant loss of the information present in the original bipartite structure. To address this we present three novel graph metrics that compute document coherence on the original bipartite graph of sentences and entities. Evaluation on standard settings shows that: (i) one of our coherence metrics beats the state of the art in terms of coherence accuracy; and (ii) all three of our coherence metrics improve retrieval effectiveness because, as closer analysis reveals, they capture aspects of document quality that go undetected by both keyword-based standard ranking and by spam filtering. This work contributes to document coherence metrics that are theoretically principled, parameter-free, and useful to IR.
29th International Conference on Legal Knowledge and Information Systems (JURIX'16)
What is a case decided by the European Court of Human Rights about?
The Courts own case database,... more What is a case decided by the European Court of Human Rights about? The Courts own case database, HUDOC , lists all the articles mentioned in a specific case in their metadata. They also supply a number of keywords, but these keywords for the most part are reduced to repeating phrases from the relevant articles. In order to enhance information retrieval about case content, without relying on manual labor and subjective judgment, we propose in this paper a quantitative method that gives a better indication of case content in terms of which articles a given case is more closely associated with. To do so, we rely on the network structure induced by existing case-to-case and case-to-article citations and propose two computational approaches (referred to as MainIn and MainOuT) which result in assigning one representative article to each case. We validate the approach by selecting a sample of important cases and comparing manual investigation of real content of those cases with the MainIn and MainOut articles. Results show that MainIn in particular is able to infer correctly the real content in most of the cases.
—Do case citations reflect the " real " importance of individual judgments for the legal system c... more —Do case citations reflect the " real " importance of individual judgments for the legal system concerned? This question has long been puzzling empirical legal scholars. Existing research typically studies case citation networks as a whole applying traditional network metrics stemming from graph theory. Those approaches are able to detect globally important cases, but since they do not take time explicitly into account, they cannot provide a comprehensive account of the dynamics behind the network structure and its evolution. In this paper we provide such a description, using two node importance metrics that take time into account to study important cases in the Court of Justice of the European Union over time. We then compare cases deemed as important by the metrics, with a set of 50 cases selected by the Court as the most important (landmark) cases. Our contribution is twofold. First, with regard to network science, we show that structural and time-related properties are complementary, and necessary to obtain a complete and nuanced picture of the citation network. Second, with regard to the case law of the Court, this study provides empirical evidence clarifying the motivation of the Court when selecting the landmark cases, revealing the importance of symbolic and historical cases in the selection. In addition, the temporal analysis sheds new light on the network properties specific to the landmark cases that distinguishes them from the rest of the cases. We validate our results by providing legal interpretations that sustain the highlights provided by the proposed network analysis.
Since the early 2000’s, the Internet topology has been an attractive and important research topic... more Since the early 2000’s, the Internet topology has been an attractive and important research topic, either for developing data collection mechanisms, and for analyzing and modeling the network. Beside traditional aspects of the Internet topology (i.e., IP interface, router, and AS levels), recent researches focused on intermediate promising visions of the topology, namely Point-of-Presence (PoP) and subnets (i.e., a set of devices that are located on the same connection medium and that can communicate directly with each other at the link layer). This paper focuses on network subnet discovery by proposing a new tool called TreeNET. One of the key aspects of TreeNET is that it builds a tree representing the way subnets are located with respect to each other. This tree allows TreeNET to obtain additional information on the network, leading to better analysis of the collected data. In this paper, we demonstrate the potential of TreeNET through the evaluation of its key algorithmic steps and the study of measurements collected from the PlanetLab testbed.
The 18th Asia-Pacific Network Operations and Management Symposium
After being the support of the data and voice
convergence, the Internet has become one of the mai... more After being the support of the data and voice convergence, the Internet has become one of the main video providers such as TV-stream. As an alternative to limited or expensive technologies, P2PTV has turned out to be a promising support for such applications. This infrastructure strongly relies on the overlay composed by the peers that consume and diffuse video contents at the same time. Understanding the dynamical properties of this overlay, and in particular how the users switch from one overlay to another, appears to be a key aspect if one wants to improve the quality of P2PTV.
In this paper, we investigate the question of relying on non- invasive measurement techniques to track the presence of users on several channels of P2PTV. Using two datasets obtained by using network measurement on P2PTV infrastructure, we show that such approach contains sufficient information to track the presence of users on several channels. Besides, exploiting the view provided by sliding time windows, we are able to refine the analysis and track users that switch from one channel to another, leading to the detection of super-peers and providing explanations of the different roles they can play in the infrastructure. In addition, by comparing the results obtained on the two datasets, we show how such analyses can shed some light on the evolution of the infrastructure polic
Many studies have proposed to apply artificial intelligence techniques to
legal networks, whether... more Many studies have proposed to apply artificial intelligence techniques to legal networks, whether it be for highlighting legal reasoning, resolving conflict or extracting information from legal databases. In this context, a new line of research has recently emerged which consists in considering legal decisions as elements of complex networks and conduct a structural analysis of the relations between the decisions. It has proved to be efficient for detecting important decisions in legal rulings. In this paper, we follow this approach and propose to extend structural analyses with temporal properties. We define in particular the notion of relative in-degree, temporal distance and average longevity and use those metrics to rank the legal decisions of the two first trials of the International Criminal Court. The results presented in this paper highlight non trivial temporal properties of those legal networks, such as the presence of decisions with an unexpected high longevity, and show the relevance of the proposed relative in-degree property to detect landmark decisions. We validate the outcomes by confronting the results to the one obtained with the standard in-degree property and provide juridical explanations of the decisions identified as important by our approach
For a long time now, researchers have worked on
defining different metrics able to characterize t... more For a long time now, researchers have worked on defining different metrics able to characterize the importance of nodes in networks. Among them, centrality measures have proved to be pertinent as they relate the position of a node in the structure to its ability to diffuse an information efficiently. The case of dynamic networks, in which nodes and links appear and disappear over time, led the community to propose extensions of those classical measures. Yet, they do not investigate the fact that the network structure evolves and that node importance may evolve accordingly. In the present paper, we propose temporal extensions of notions of centrality, which take into account the paths existing at any given time, in order to study the time evolution of nodes’ importance in dynamic networks. We apply this to two datasets and show that the importance of nodes does indeed vary greatly with time. We also show that in some cases it might be meaningless to try to identify nodes that are consistently important over time, thus strengthening the interest of temporal extensions of centrality measures.
2014 Ieee 22nd International Symposium on Modelling Analysis Simulation of Computer and Telecommunication Systems, Sep 9, 2014
The classical approach for Internet topology measurement consists in distributively collecting as... more The classical approach for Internet topology measurement consists in distributively collecting as much data as possible and merging it into one single piece of topology on which are conducted subsequent analysis. Although this approach may seem reasonable, in most cases network measurements performed in this way suffer from some or all of the following limitations: they give only partial views of the networks under concern, these views may be intrinsically biased, and they contain erroneous data due to the measurement tools. Here we present a new tool, named UDP Ping , that relies on a very different approach for the measurement of the Internet topology. Its basic principle is to measure the interface of a given target directed toward a monitor which sends the measurement probe. We demonstrate how to use it to deploy real world-wide measurements that provide reliable (i.e. bias and error free) knowledge of the Internet topology, namely the degree distribution of routers in the core Internet in our example.
Many real-world networks lend themselves to the use of graphs
for analysing and modelling their s... more Many real-world networks lend themselves to the use of graphs for analysing and modelling their structure. But such a simple representation has proven to miss some important and non trivial properties hidden in the bipartite structure of the networks. Recent papers have shown that overlapping properties seem to be present in bipartite networks and that it could explain better the properties observed in simple graphs. This work intends to investigate this question by studying two proposed metrics to account for overlapping structures in bipartite networks. The study, conducted on four dataset stemming from very different contexts (computer science, juridical science and social science), shows that the most popular metrics, the clustering coefficient, turns out to be less relevant that the recent redundancy coefficient to analyse intricate overlapping properties of real networks.
ABSTRACT The development of wireless devices led the scientific community to focus more and more ... more ABSTRACT The development of wireless devices led the scientific community to focus more and more on systems of interaction composed of moving entities. In this context, different models have been proposed in an attempt to capture properties of the observed dynamics. Among those models, the edge-Markovian evolving graph model is appealing since it enables to highlight temporal dependencies in the evolution of the graphs. This model relies on two parameters accounting respectively for the creation and suppression of links in the graph. Thus it assumes that these two parameters are sufficient to characterise the dynamics during all the evolution of the graph. In this paper, we test this hypothesis by confronting the model to 6 datasets recording real traces of evolving networks. In particular, we study the proportion of created and deleted links over the time. The results show that 5 of the 6 case studies present an heterogeneous distribution of those fractions which contradicts the underlying hypothesis of the model. Besides, in order to understand the importance this might have as regard structural properties of real networks, we also study the impact the Markovian model has on the mean degree over the time. It turns out that even in the suitable case, the model fails to reproduce correctly this property which indicates its inadequacy for even more complex properties of real evolving networks.
The use of process calculi to represent biological systems has led to the design of different for... more The use of process calculi to represent biological systems has led to the design of different formalisms such as brane calculi and κ-calculus. Both have proved to be useful to model different types of biological systems. As an attempt to unify the formalisms, we introduce the bioκ-calculus , a simple calculus for describing proteins and cells, in which bonds are represented by means of shared names and interactions are modelled at the domain level. In bioκ-calculus, protein-protein interactions have to be at most binary and cell interactions have to fit with sort constraints. In this contribution we define the semantics of bioκ-calculus, analyse its properties, discuss the expressivity of the calculus by modelling two significant examples -- a signalling pathway and a virus infection --, and study an implementation in Milner’s π-calculus.
Proceedings of the First International Conference on Mechanisms Symbols and Models Underlying Cognition Interplay Between Natural and Artificial Computation Volume Part I, 2005
The Cologne-Twente Workshop (CTW) on Graphs and Combinatorial Optimization started off as a serie... more The Cologne-Twente Workshop (CTW) on Graphs and Combinatorial Optimization started off as a series of workshops organized bi-annually by either Köln University or Twente University. As its importance grew over time, it re-centered its geographical focus by including northern Italy (CTW04 in Menaggio, on the lake Como and CTW08 in Gargnano, on the Garda lake). This year, CTW (in its eighth edition) will be staged in France for the first time: more precisely in the heart of Paris, at the Conservatoire National d’Arts et Métiers (CNAM), between 2nd and 4th June 2009, by a mixed organizing committee with members from LIX, École Polytechnique and CEDRIC, CNAM. As tradition warrants, a special issue of Discrete Applied Mathematics (DAM) will be devoted to CTW09, containing full-length versions of selected presentations given at the workshop and possibly other contributions related to the workshop topics. The deadline for submission to this issue will be posted in due time on the
... presse au moment o`u j'écris ces lignes, Mau-reen, Assia, Germain, Daniel, Michel (ça, c&... more ... presse au moment o`u j'écris ces lignes, Mau-reen, Assia, Germain, Daniel, Michel (ça, c'est fait!), Erwan pour sa dignité dans sa défaite au championnat de raclette, Julien, Géraldine, Milan and Co, Jean qui s'est révélé être un collaborateur en plus d'un pote, David dont j'at ...
The modern world, especially in medicine, in the social sphere and in the environment, increasing... more The modern world, especially in medicine, in the social sphere and in the environment, increasingly depends on or confronts very large systems comprised of many interacting entities. The data collected on such systems, typically on enormous scales, poses formidable challenges for efforts to reconstruct their multiscale dynamics and their multiple downward and upward influences. The task requires the help of a formal epistemology and massive computation, and a generalization of the kind of" open science" originally ...
Since 1989, no major European merger has been able to go through without EU approval. The introdu... more Since 1989, no major European merger has been able to go through without EU approval. The introduction of a centralized merger control procedure was another increase in the powers of the Commission's Directorate‐General for Competition (DG COMP). While some see it playing a neo‐mercantilist role in a positive European integration, others underline its neoliberal ideological roots. Through our analysis of all merger decisions made between 1990 and 2016 (6,161 cases), we instead find evidence for market‐centred negative integration: DG COMP is particularly harsh towards coordinated market economies and targets sectors that have high levels of state intervention, thus thwarting the rise of ‘European champions’. Our interviews with merger experts and the decision citation data further suggest that this market‐centred logic of enforcement is not necessarily driven by ideology, but by the silent logic of bureaucratic autonomy. We thus contribute to the debate on the EU as a supranatio...
Since the early 2000’s, the Internet topology has been an attractive and important research topic... more Since the early 2000’s, the Internet topology has been an attractive and important research topic, either for developing data collection mechanisms, and for analyzing and modeling the network. Beside traditional aspects of the Internet topology (i.e., IP interface, router, and AS levels), recent researches focused on intermediate promising visions of the topology, namely Point-of-Presence (PoP) and subnets (i.e., a set of devices that are located on the same connection medium and that can communicate directly with each other at the link layer). This paper focuses on network subnet discovery by proposing a new tool called TreeNET. One of the key aspects of TreeNET is that it builds a tree representing the way subnets are located with respect to each other. This tree allows TreeNET to obtain additional information on the network, leading to better analysis of the collected data. In this paper, we demonstrate the potential of TreeNET through the evaluation of its key algorithmic steps ...
What is a case decided by the European Court of Human Rights about? The Courts own case database,... more What is a case decided by the European Court of Human Rights about? The Courts own case database, HUDOC , lists all the articles mentioned in a specific case in their metadata. They also supply a number of keywords, but these keywords for the most part are reduced to repeating phrases from the relevant articles. In order to enhance information retrieval about case content, without relying on manual labor and subjective judgement, we propose in this paper a quantitative method that gives a better indication of case content in terms of which articles a given case is more closely associated with. To do so, we rely on the network structure induced by existing case-to-case and case-to-article citations and propose two computational approaches (referred to as MainIn and MainOut) which result in assigning one representative article to each case. We validate the approach by selecting a sample of important cases and comparing manual investigation of real content of those cases with the MainI...
The ability to detect important nodes in temporal networks has been investigated lately. This has... more The ability to detect important nodes in temporal networks has been investigated lately. This has been a challenge on both the theoretical aspects as well as computational ones. In this study we propose and evaluate different strategies to detect nodes that have high temporal closeness.
Many studies have proposed to apply artificial intelligence techniques to legal networks, whether... more Many studies have proposed to apply artificial intelligence techniques to legal networks, whether it be for highlighting legal reasoning, resolving conflict or extracting information from legal databases. In this context, a new line of research has recently emerged which consists in considering legal decisions as elements of complex networks and conduct a structural analysis of the relations between the decisions. It has proved to be efficient for detecting important decisions in legal rulings. In this paper, we follow this approach and propose to extend structural analyses with temporal properties. We define in particular the notion of relative in-degree, temporal distance and average longevity and use those metrics to rank the legal decisions of the two first trials of the International Criminal Court. The results presented in this paper highlight non trivial temporal properties of those legal networks, such as the presence of decisions with an unexpected high longevity, and show ...
While the ever-increasing emergence of online services has led to a growing interest in the devel... more While the ever-increasing emergence of online services has led to a growing interest in the development of recommender systems, the algorithms underpinning such systems have begun to be criticized for their role in limiting the variety of content exposed to users. In this context, the notion of diversity has been proposed as a way of mitigating the side effects resulting from the specialization of recommender systems. In this paper, using a well-known recommender system that makes use of collaborative filtering in the context of musical content, we analyze the diversity of recommendations generated through the lens of the recently proposed information network diversity measure. The results of our study offer significant insights into the effect of algorithmic recommendations. On the one hand, we show that the musical selections of a large proportion of users are diversified as a result of the recommendations. On the other hand, however, such improvements do not benefit all users. Th...
Uploads
documents according to both their relevance and their coherence has been shown to increase retrieval effectiveness [34, 37].
The state of the art in unsupervised coherence modelling represents documents as bipartite graphs of sentences and discourse entities, and then projects these bipartite graphs into one–mode undirected graphs. However, one–mode projections may incur significant loss of the information present in the original bipartite structure. To address this we present three novel graph metrics that compute document coherence
on the original bipartite graph of sentences and entities. Evaluation on standard settings shows that: (i) one of our coherence metrics beats the state of the art in terms of coherence accuracy; and (ii) all three of our coherence metrics improve
retrieval effectiveness because, as closer analysis reveals, they capture aspects of document quality that go undetected by both keyword-based standard ranking and by spam filtering. This work contributes to document coherence metrics that are
theoretically principled, parameter-free, and useful to IR.
The Courts own case database, HUDOC , lists all the articles mentioned in a specific case in their metadata. They also supply a number of keywords, but these keywords for the most part are reduced to repeating phrases from the relevant articles. In order to enhance information retrieval about case content, without relying on manual labor and subjective judgment, we propose in this paper a quantitative method that
gives a better indication of case content in terms of which articles a given case is more closely associated with. To do so, we rely on the network structure induced by existing case-to-case and case-to-article citations and propose two computational approaches (referred to as MainIn and MainOuT) which result in assigning one representative article to each case. We validate the approach by selecting a sample of important cases and comparing manual investigation of real content of those cases with the MainIn and MainOut articles. Results show that MainIn in particular is able to infer correctly the real content in most of the cases.
This paper focuses on network subnet discovery by proposing a new tool called TreeNET. One of the key aspects of TreeNET is that it builds a tree representing the way subnets are located with respect to each other. This tree allows TreeNET to obtain additional information on the network, leading to better analysis of the collected data. In this paper, we demonstrate the potential of TreeNET through the evaluation of its key algorithmic steps and the study of measurements collected from the PlanetLab testbed.
convergence, the Internet has become one of the main video
providers such as TV-stream. As an alternative to limited or
expensive technologies, P2PTV has turned out to be a promising
support for such applications. This infrastructure strongly relies
on the overlay composed by the peers that consume and diffuse
video contents at the same time. Understanding the dynamical
properties of this overlay, and in particular how the users switch
from one overlay to another, appears to be a key aspect if one
wants to improve the quality of P2PTV.
In this paper, we investigate the question of relying on non-
invasive measurement techniques to track the presence of users
on several channels of P2PTV. Using two datasets obtained by
using network measurement on P2PTV infrastructure, we show
that such approach contains sufficient information to track the
presence of users on several channels. Besides, exploiting the
view provided by sliding time windows, we are able to refine the
analysis and track users that switch from one channel to another,
leading to the detection of super-peers and providing explanations
of the different roles they can play in the infrastructure. In
addition, by comparing the results obtained on the two datasets,
we show how such analyses can shed some light on the evolution
of the infrastructure polic
legal networks, whether it be for highlighting legal reasoning, resolving conflict or extracting information from legal databases. In this context, a new line of research has recently emerged which consists in considering legal decisions as elements of complex networks and conduct a structural analysis of the relations between the decisions. It has proved to be efficient for detecting important decisions in legal rulings. In this paper, we follow this approach and propose to extend structural
analyses with temporal properties. We define in particular the notion of relative in-degree, temporal distance and average longevity and use those metrics to rank the legal decisions of the two first trials of the International Criminal Court. The results presented in this paper highlight non trivial temporal properties of those legal networks, such as the presence of decisions with an unexpected high longevity, and show the relevance of the proposed relative in-degree property to detect landmark decisions. We validate the outcomes by confronting the results to the one obtained with the standard in-degree property and provide juridical explanations of the decisions identified as important by our approach
defining different metrics able to characterize the importance
of nodes in networks. Among them, centrality measures have
proved to be pertinent as they relate the position of a node in the
structure to its ability to diffuse an information efficiently. The
case of dynamic networks, in which nodes and links appear and
disappear over time, led the community to propose extensions of
those classical measures. Yet, they do not investigate the fact that
the network structure evolves and that node importance may
evolve accordingly. In the present paper, we propose temporal
extensions of notions of centrality, which take into account the
paths existing at any given time, in order to study the time
evolution of nodes’ importance in dynamic networks. We apply
this to two datasets and show that the importance of nodes does
indeed vary greatly with time. We also show that in some cases it
might be meaningless to try to identify nodes that are consistently
important over time, thus strengthening the interest of temporal
extensions of centrality measures.
for analysing and modelling their structure. But such a simple representation has proven to miss some important and non trivial properties hidden in the bipartite structure of the networks. Recent papers have shown that overlapping properties seem to be present in bipartite networks and that it could explain better the properties observed in simple graphs. This work intends to investigate this question by studying two proposed metrics to account for overlapping structures in bipartite networks. The study, conducted on four dataset stemming from very different contexts (computer science, juridical science and social science), shows that the most popular metrics, the clustering coefficient, turns out to be less relevant that the recent redundancy coefficient to analyse intricate overlapping properties of real networks.
documents according to both their relevance and their coherence has been shown to increase retrieval effectiveness [34, 37].
The state of the art in unsupervised coherence modelling represents documents as bipartite graphs of sentences and discourse entities, and then projects these bipartite graphs into one–mode undirected graphs. However, one–mode projections may incur significant loss of the information present in the original bipartite structure. To address this we present three novel graph metrics that compute document coherence
on the original bipartite graph of sentences and entities. Evaluation on standard settings shows that: (i) one of our coherence metrics beats the state of the art in terms of coherence accuracy; and (ii) all three of our coherence metrics improve
retrieval effectiveness because, as closer analysis reveals, they capture aspects of document quality that go undetected by both keyword-based standard ranking and by spam filtering. This work contributes to document coherence metrics that are
theoretically principled, parameter-free, and useful to IR.
The Courts own case database, HUDOC , lists all the articles mentioned in a specific case in their metadata. They also supply a number of keywords, but these keywords for the most part are reduced to repeating phrases from the relevant articles. In order to enhance information retrieval about case content, without relying on manual labor and subjective judgment, we propose in this paper a quantitative method that
gives a better indication of case content in terms of which articles a given case is more closely associated with. To do so, we rely on the network structure induced by existing case-to-case and case-to-article citations and propose two computational approaches (referred to as MainIn and MainOuT) which result in assigning one representative article to each case. We validate the approach by selecting a sample of important cases and comparing manual investigation of real content of those cases with the MainIn and MainOut articles. Results show that MainIn in particular is able to infer correctly the real content in most of the cases.
This paper focuses on network subnet discovery by proposing a new tool called TreeNET. One of the key aspects of TreeNET is that it builds a tree representing the way subnets are located with respect to each other. This tree allows TreeNET to obtain additional information on the network, leading to better analysis of the collected data. In this paper, we demonstrate the potential of TreeNET through the evaluation of its key algorithmic steps and the study of measurements collected from the PlanetLab testbed.
convergence, the Internet has become one of the main video
providers such as TV-stream. As an alternative to limited or
expensive technologies, P2PTV has turned out to be a promising
support for such applications. This infrastructure strongly relies
on the overlay composed by the peers that consume and diffuse
video contents at the same time. Understanding the dynamical
properties of this overlay, and in particular how the users switch
from one overlay to another, appears to be a key aspect if one
wants to improve the quality of P2PTV.
In this paper, we investigate the question of relying on non-
invasive measurement techniques to track the presence of users
on several channels of P2PTV. Using two datasets obtained by
using network measurement on P2PTV infrastructure, we show
that such approach contains sufficient information to track the
presence of users on several channels. Besides, exploiting the
view provided by sliding time windows, we are able to refine the
analysis and track users that switch from one channel to another,
leading to the detection of super-peers and providing explanations
of the different roles they can play in the infrastructure. In
addition, by comparing the results obtained on the two datasets,
we show how such analyses can shed some light on the evolution
of the infrastructure polic
legal networks, whether it be for highlighting legal reasoning, resolving conflict or extracting information from legal databases. In this context, a new line of research has recently emerged which consists in considering legal decisions as elements of complex networks and conduct a structural analysis of the relations between the decisions. It has proved to be efficient for detecting important decisions in legal rulings. In this paper, we follow this approach and propose to extend structural
analyses with temporal properties. We define in particular the notion of relative in-degree, temporal distance and average longevity and use those metrics to rank the legal decisions of the two first trials of the International Criminal Court. The results presented in this paper highlight non trivial temporal properties of those legal networks, such as the presence of decisions with an unexpected high longevity, and show the relevance of the proposed relative in-degree property to detect landmark decisions. We validate the outcomes by confronting the results to the one obtained with the standard in-degree property and provide juridical explanations of the decisions identified as important by our approach
defining different metrics able to characterize the importance
of nodes in networks. Among them, centrality measures have
proved to be pertinent as they relate the position of a node in the
structure to its ability to diffuse an information efficiently. The
case of dynamic networks, in which nodes and links appear and
disappear over time, led the community to propose extensions of
those classical measures. Yet, they do not investigate the fact that
the network structure evolves and that node importance may
evolve accordingly. In the present paper, we propose temporal
extensions of notions of centrality, which take into account the
paths existing at any given time, in order to study the time
evolution of nodes’ importance in dynamic networks. We apply
this to two datasets and show that the importance of nodes does
indeed vary greatly with time. We also show that in some cases it
might be meaningless to try to identify nodes that are consistently
important over time, thus strengthening the interest of temporal
extensions of centrality measures.
for analysing and modelling their structure. But such a simple representation has proven to miss some important and non trivial properties hidden in the bipartite structure of the networks. Recent papers have shown that overlapping properties seem to be present in bipartite networks and that it could explain better the properties observed in simple graphs. This work intends to investigate this question by studying two proposed metrics to account for overlapping structures in bipartite networks. The study, conducted on four dataset stemming from very different contexts (computer science, juridical science and social science), shows that the most popular metrics, the clustering coefficient, turns out to be less relevant that the recent redundancy coefficient to analyse intricate overlapping properties of real networks.
In this contribution we define the semantics of bioκ-calculus, analyse its properties, discuss the expressivity of the calculus by modelling two significant examples -- a signalling pathway and a virus infection --, and study an implementation in Milner’s π-calculus.