Maria Malek

A multi-agent system for collaborative bookmarking

In this paper we describe a new distributed collaborative bookmark system, called COWING (for COllaborative Web IndexING system). The COWING system is composed of a set of assistant agents, called WINGS, and a central agent that manages the users organization. Each user is assisted by a Wing agent that performs two tasks: learning the users strategy in classifying her/his own bookmarks and interacting with other WING agents in order to fetch new bookmarks that match the local user information need.

Toward ontology-based personalization of a recommender system in social network

Personalized search, navigation and content delivery techniques have attracted interest in the recommender systems as a means to decrease search ambiguity and return results most relevant to a particular user preferences. In this paper, we study the effect of incorporating user semantic profile derived from past users behavior and preferences on the accuracy of a recommender system. We present a preliminary work which aims at tackling the most technical issues due to the integration of an ontology-based semantic user profile within a hybrid recommender system based on our early released guided recommender algorithm. A semantic user profile context is represented as an instance of a reference domain ontology in which concepts are annotated by interest scores.

Searching Frequent Itemsets by Clustering Data: Towards a Parallel Approach Using Mapreduce

We propose a new algorithm for searching frequent itemsets in large data bases. The idea is to start searching from a set of representative examples instead of testing the 1-itemset,the k-itemset and so on. A clustering algorithm is firstly applied in order to cluster the transactions into k clusters. The set of the k representative examples will be used as the starting point for searching frequent itemsets. Each cluster is represented by the most representative example. We show some preliminary results and we then propose a parallel version of this algorithm based on the MapReduce Framework.

CoWing: A Collaborative Bookmark Management System

In this paper we describe a new distributed collaborative bookmark system, called COWING (for COllaborative Web IndexING system). The goal of the COWING system is to provide a group of organized users with a computerized-support that enable them to share their experiences in managing their bookmark repositories. The COWING system is composed of a set of assistant agents, called WINGS, and a central agent that manages the users organization. A WING agent assists each user. This agent performs two main tasks: learning the users strategy in classifying her/his bookmarks and interacting with other WING agents in order to fetch new bookmarks that match the local user information need.

Computing social networks for information sharing: a case-based approach

In this paper we describe a peer-to-peer approach that ails at allowing a group of like-minded people to share relevant documents in an implicit way. We suppose that user save their documents in a local user-defined hierarchy. the association between documents and hierarchy nodes (or folders) is used by a supervised hybrid neural-CBR classifier in order to learn the user classification strategy. This strategy is then used to compute correlations between local folders and remote ones allowing to recommend documents without having a shared hierarchy. Another CBR system is used to memorize how good queries are answered by peer agents allowing to learn a dynamic community of peer agents to be associated with each local folder.

Parallel and distributed core label propagation with graph coloring: Parallel and distributed core label propagation with graph coloring

Label propagation is one of the fastest methods for community detection with near linear time complexity. It is a local method where each node interacts with its neighbors to change its own label. Unfortunately, it has two major drawbacks. The first is a bad propagation, sometimes leading to huge communities without meaning (the giant communities problem). The second is related to its instability. Trials of a label propagation algorithm rarely give the same result. We propose to use a more stable variant of label propagation with a core method attached in order to obtain a more deterministic algorithm. This implementation will be done in a parallel and distributed environment on Hadoop using the MapReduce framework in order to apply this method with graphs having millions of nodes and edges. The main contribution of this paper is to model a parallel and distributed algorithm to achieve this purpose. A case study of the algorithm proposed is described at the end of the article along with the comparison of our results with other well‐known algorithms.

A New Label Propagation With Dams

Label propagation is one of the fastest methods for community detection, with a near linear time complexity. Its a local method, where each node interacts with its neighbourhood to change its own label. Unfortunately, this method has two major drawbacks. The first is a bad propagation which can lead to obtain huge communities without sense (monster communities problem). The second is the instability of the method. Each trial of a label propagation algorithm gives rarely the same result. In this paper, we propose to do a study on the label propagation by putting artificial dams on edges of some networks in order to limit the propagation and to observe if this can lead to better results. We then apply an existing method based on a co-occurrence frequency matrix in order to stabilise the algorithm.

Overlapping Community Detection Using Core Label Propagation and Belonging Function

Label propagation is one of the fastest methods for community detection, with a near linear time complexity. It acts locally. Each node interacts with neighbours to change its own label by a majority vote. But this method has three major drawbacks: (i) it can lead to huge communities without sense called also monster communities, (ii) it is unstable, and (iii) it is unable to detect overlapping communities.

Graph Searching Algorithms for Semantic-Social Recommendation

In this paper we present two recommendation algorithms, called Node-Edge-Based and Node-Based recommendation algorithms. These algorithms are designed to recommend items to users connected via social network. Our algorithms are based on three main features: a social network analysis measure (degree centrality), the graph searching algorithm (Depth First Search algorithm), and the semantic similarity measure (which measures the closeness between the input item and users). We apply these algorithms to a real dataset (Amazon dataset) and we compare them with item-based collaborative filtering and hybrid recommendation algorithms. Our results show good precision as well as in a good performance in terms of runtime. Moreover, Node-Edge-Based and Node-Based algorithms search a small part of the dataset, compared to item-based and hybrid recommendation algorithms.

Exploiting Semantic and Social Information in Recommendation Algorithms

In this paper we present algorithms for recommender systems. Our algorithms rely on a semantic relevance measure and a social network analysis measure to partially explore the network using depth-first search and breath-first search strategies. We apply these algorithms to a real data set and we compare them with item-based collaborative filtering and hybrid recommendation algorithms. Our experiments show that our algorithms outperform existing recommendation algorithms, while providing good precision and F-measure results.