iSWAG 2015

iSWAG Symposium

The International Symposium on Web Algorithms (iSWAG) is dedicated to all academic and industrial researchers working on algorithmic problems related to the web. The aim of iSWAG 2015 is to cover as completely as possible the field of research on algorithms for solving web related problems.

Accepted posters

Time-aware trust model for recommender systems

Charif Haydar, Boyer Anne and Roussanaly Azim

Trust is an imperative issue in any human society. It is built up with the survey of recurrent interactions between fellows. By consequence, trust is sensible to the time, which we call the temporal factor is trust relationship. During the last decade, the arise of social web resulted a serious need to a trust model for this virtual society. Many models were proposed to represent computational trust in different applications of social web. Even models that represent trust as incremental measurement, do not accord enough importance to the time axe. In this paper, we propose and compare many hypothesis to integrate the temporal factor in measuring trust between fellows.

paper     poster

Information Retrieval enhancement using Social Networks

Bougchiche Lilia

The aim of this paper is to present a preliminary work about the enhance of information retrieval systems by exploring tags coming from social networks. We discuss how to obtain folksonomies (social indexing) and their use on a search approach that models tags assigned to documents. For this, a document is represented by formal concept then we construct a bipartite graph. From this later, we retrieve emergent tags that describe better a document.


Skill-Aware Task Assignment in Crowdsourcing Applications

Panagiotis Mavridis, David Gross-Amblard and Zoltan Miklos

Besides simple human intelligence tasks such as image labeling, crowdsourcing platforms propose more and more tasks that require very specific skills. In such a setting we need to model skills that are required to execute a particular job. At the same time in order to match tasks to the crowd, we have to model the expertise of the participants. We present such a skill model that relies on a taxonomy. We also introduce task assignment algorithms to optimize the result quality. We illustrate the effectiveness of our algorithms and models through preliminary experiments with synthetic datasets.

paper    poster

Reordering Very Large Graphs for Fun & Profit

Lionel Auroux, Marwan Burelle and Robert Erra

We have made experiments with reordering algorithms on sparse very large graphs (VLGs), we have considered only undirected, unweighted, sparse and huge graphs, i.e. G = (V, E) with n = |V | from million to billion of nodes and with |E| = O(|V |). The problem of reordering a matrix to enhance the computation time (and sometimes the memory) is traditional in numerical algorithms but we focus on this short paper on results obtained for the approximate computation of the diameter of a sparse VLG (with some graphs on various different computers). The problem of reordering a graph has already been pointed, explicitly or implicitly by a lot of people, from the numerical community but also from the graph community, like the authors of the Louvain algorithm when they write that choosing an order is thus worth studying since it could accelerate the Louvain algorithm. Our experimental results show clearly that it can be worth (and simple) to preprocess a sparse VLG with a reordering algorithm.


Temporal Reconciliation Based on Entity Information

Paul Martin, Marc Spaniol and Antoine Doucet

Temporal classification of Web contents requires a “notion” about them. This is particularly relevant when contents contain several dates and a human “interpretation” is required in order to chose the appropriate time point. The dating challenge becomes even more complex, when images have to be dated based on the content describing them. In this paper, we present a novel timestamping approach based on semantics derived from the document. To this end, we will first introduce our experimental dataset and then explain our temporal reconciliation pipeline. In particular, we will explain the process of temporal reconciliation by incorporating information derived from named entities.

paper    poster

ThemaMap: a Free Versatile Data Analysis and Visualization Tool

Jacques Madelaine, Gaétan Richard and Gilles Domalain

The ThemaMap project aims to provide a friendly and versatile thematic cartographic tool available as an OpenSource free software. It allows various classic and fancy data visualizations. Data table manipulation allows simple statistics indicators computing or data (alphanumeric and geographic) grouping and filtering. This paper explains how it can be a good candidate tool to make Open Data accessible to everyone.

paper    poster

A comparison of graph clustering algorithms

Jean Creusefond

The community detection problem is very natural : given a set of people and their relationships, can we understand the underlying structure of social groups? The applications are numerous in marketing, politics, social statistics, ... Thanks to technological improvements and change of uses, many social networks of various sizes has been automatically extracted from recorded social relationships. The most obvious examples are the social network websites, but other networks are also studied as social networks : collaboration between scientists, who-talks-to-whom on the phone/on emails, etc. Automatic extraction made large networks available for study, and the community detection algorithms that we could evaluate with ease by watching the result on small instances before, can not be compared on real-world networks. We therefore try to find common ground among the various clustering algorithms. Indeed, most of them share design similarities, as the underlying assumptions about the characteristics of communities or the general steps of the algorithm. Our experiments show to what extent these similarities imply similarities of results.

paper    poster

Algorithms for continuous top-k processing in social networks

Abdulhafiz Alkhouli, Dan Vodislav and Boris Borzic

Information streams are mainly produced today by social networks, but current methods for continuous top-k processing of such streams are still limited to contentbased similarity. We present the SANTA algorithm, able to handle also social network criteria and events, and report a preliminary comparison with an extension of a state-of-the-art algorithm.


First experiments with Parametrized Neural Network Language Models for Information Retrieval

Nicolas Desprès, Sylvain Lamprier and Benjamin Piwowarski

Information Retrieval (IR) models need to deal with two difficult issues, vocabulary mismatch and term dependencies. A multitude of solutions has been proposed to solve each of these two problems, but no principled model solve both. In parallel, in the last few years, language models based on neural networks have been used to cope with complex natural language processing tasks like emotion and paraphrase detection. Although they present good abilities to cope with both term dependencies and vocabulary mismatch problems, thanks to the distributed representation of words they are based upon, such models could not be used readily in IR, where the estimation of one language model per document (or query) is required. This is both computationally unfeasible and prone to over-fitting. Within the language model IR framework, we propose and study the use of a generic language model as well as a document-specific language model..