Mourad Ouziri

A penalty-based approach for qos dissatisfaction using fuzzy rules

Quality of Service (QoS) guarantees are commonly defined in Service Level Agreements (SLAs) between provider and consumer of services. Such guarantees are often violated due to various reasons. QoS violation requires a service adaptation and penalties have to be associated when promises are not met. However, there is a lack of research in defining and assessing penalties according to the degree of violation. In this paper, we provide an approach based on fuzzy logic for modelling and measuring penalties with respect to the extent of QoS violation. Penalties are assigned by means of fuzzy rules.

Semantic Indexing for Intelligent Browsing of Distributed Data

We present in this paper a semantic indexing technics based on description logics. Data to be indexed are semantically organized as a Topic Map. The index is constructed according to data organization, user profile, and data distribution (association rules). This way leads to obtain a more efficient index and represents more semantics. The index is well adapted to jointly query and navigate in the topic map. DL allows to represent semantics and performs powerful reasoning. The index structure is based on subsumption relationships (for intra-concept indexing) and roles (for inter-concepts indexing).

Fusion of Big RDF Data: A Semantic Entity Resolution and Query Rewriting-Based Inference Approach

This paper presents an efficient approach to query big RDF datasources in order to get more relevant and complete results. The approach deals with two important heterogeneities in huge amount of data: semantic and URI-based entity identification heterogeneities. The paper proposes: (1) a semantic entity resolution approach based on inference mechanism to manage ambiguity of real world entities for linking data at the semantic and URI levels (2) a MapReduce-based query rewriting approach based on entity resolution results to include implicit data into query results (3) algorithms based on MapReduce paradigm to deal with huge amounts of data.

Discovering Best Teams for Data Leak-Aware Crowdsourcing in Social Networks

Crowdsourcing is emerging as a powerful paradigm to help perform a wide range of tedious tasks in various enterprise applications. As such applications become more complex, crowdsourcing systems often require the collaboration of several experts connected through professional/social networks and organized in various teams. For instance, a well-known car manufacturer asked fans to contribute ideas for the kinds of technologies that should be incorporated into one of its cars. For that purpose, fans needed to collaborate and form teams competing with each others to come up with the best ideas. However, once teams are formed, each one would like to provide the best solution and treat that solution as a “trade secret,” hence preventing any data leak to its competitors (i.e., the other teams). In this article, we propose a data leak--aware crowdsourcing system called SocialCrowd. We introduce a clustering algorithm that uses social relationships between crowd workers to discover all possible teams while avoiding interteam data leakage. We also define a ranking mechanism to select the “best” team configurations. Our mechanism is based on the semiring approach defined in the area of soft constraints programming. Finally, we present experiments to assess the efficiency of the proposed approach.

Quality-Based Online Data Reconciliation

One of the main challenges in data matching and data cleaning, in highly integrated systems, is duplicates detection. While the literature abounds of approaches detecting duplicates corresponding to the same real-world entity, most of these approaches tend to eliminate duplicates (wrong information) from the sources, hence leading to what is called data repair. In this article, we propose a framework that automatically detects duplicates at query time and effectively identifies the consistent version of the data, while keeping inconsistent data in the sources. Our framework uses matching dependencies (MDs) to detect duplicates through the concept of data reconciliation rules (DRR) and conditional function dependencies (CFDs) to assess the quality of different attribute values. We also build a duplicate reconciliation index (DRI), based on clusters of duplicates detected by a set of DRRs to speed up the online data reconciliation process. Our experiments of a real-world data collection show the efficiency and effectiveness of our framework.

CrowdMD: Crowdsourcing-based approach for deduplication

Matching dependencies (MDs) were recently introduced as quality rules for data cleaning and entity resolution. They are rules that specify what values should be considered duplicates, and have to be matched. Defining such quality rules on a database instance, is a very expensive and a time consuming process, and requires huge efforts to analyse the whole database. In this demo paper, we present CrowdMD, a hybrid machine-crowd system for generating MDs. It first asks the crowd to determine whether a given pair, from training sample pairs, match or not. Then, it uses data mining techniques to generate attributes constituting an MD. Using a Restaurant database, we will show how the crowders can help to generate MDs by labelling the training sample through the CrowdMD user interface and how MDs can be mined from this training set.

Be a Collaborator and a Competitor in Crowdsourcing System

Crowd sourcing is emerging as a powerful paradigm to solve a wide range of tedious and complex problems in various enterprise applications. It spawns the issue of finding the unknown collaborative and competitive group of solvers. The formation of collaborative team should provide the best solution and treat that solution as a trade secret avoiding data leak between competitive teams due to reward behind the outsourcing of the issue. The formation of effective competitive teams not only requires adequate skills between members of each team, but also good team connectivity through social network and to provide the best solution and treat that solution as a trade secret avoiding data leak between teams due to reward behind the outsourcing of the issue. In this paper, we propose a data leak aware crowd sourcing system called Social Crowd. We introduce a clustering algorithm that uses social relationships between crowd workers to discover all possible teams while avoiding inter-team data leakage.

A tensor-based distributed discovery of missing association rules on the cloud

An increasing number of data applications such as monitoring weather data, data streaming, data web logs, and cloud data, are going online and are playing vital in our every-day life. The underlying data of such applications change very frequently, especially in the cloud environment. Many interesting events can be detected by discovering such data from different distributed sources and analyzing it for specific purposes (e.g., car accident detection or market analysis). However, several isolated events could be erroneous due to the fact that important data sets are either discarded or improperly analyzed as they contain missing data. Such events therefore need to be monitored globally and be detected jointly in order to understand their patterns and correlated relationships. In the context of current cloud computing infrastructure, no solutions exist for enabling the correlations between multi-source events in the presence of missing data. This paper addresses the problem of capturing the underlying latent structure of the data with missing entries based on association rules. This necessitate to factorize the data set with missing data. The paper proposes a novel model to handle high amount of data in cloud environment. It is a model of aggregated data that are confidences of association rules. We first propose a method to discover the association rules locally on each node of a cloud in the presence of missing rules. Afterward, we provide a tensor based model to perform a global correlation between all the local models of each node of the network. The proposed approach based on tensor decomposition, deals with a multi modal network where missing association rules are detected and their confidences are approximated. The approach is scalable in terms of factorizing multi-way arrays (i.e. tensor) in the presence of missing association rules. It is validated through experimental results which show its significance and viability in terms of detecting missing rules.

Accessing the Distributed Learner Profile in the Semantic Web

Adaptive e-learning systems need to get complete learner profile to make efficient personalization. However, learner profile is dispersed over multiple heterogeneous e-learning systems. Unfortunately, these e-learning systems are heterogeneous what makes difficult to get complete learner profile. By using semantic web technology, namely Topic Maps, we show how we perform integration of heterogeneous fragments of learner profile to get more complete one. However, this technology does not consider constraints and is not able to make reasoning. So we use, together with Topic Maps, Description Logics to represent constraints and make reasoning over integrated data.

Domed: Semantic data integration and navigation in Web-based medical records.

Medical data are stored on multiple health information systems which are heterogeneous and non-communicating. These medical information systems are often built with Web-based pages. The medical record of a patient is therefore dispatched between all these removed systems. It is then difficult to get a complete and consistent long-life medical record due to semantic and structural heterogeneities. Our aim is to propose a user interface in which the patients medical records rebuilt by the end-user himself in a simple interface where concepts are chosen in a list and linked automatically together. Therefore, the user can navigate in this space of concepts to obtain information he needs, as easily as in a web site.