Search

Collaborative filtering approach is widely used in the area of recommender systems. In fact, its predictive accuracy is supported by a large amount of additional information available on the intern...

In the last decade, social-based recommender systems have become the best way to resolve a user's cold start problem. In fact, it enriches the user's model by adding additional information provided from his social network. Most of those approaches are based on a collaborative filtering and compute similarities between the users. The authors' preliminary objective in this work is to propose an innovative context aware metric between users (called contextual influencer user). These new similarities are called C-COS, C-PCC and C-MSD, where C refers to the category. The contextual influencer user model is integrated into a social based recommendation system. The category of the items is considered as the most pertinent context element. The authors' proposal is implemented and tested within the food dataset. The experimentation proved that the contextual influencer user measure achieves 0.873, 0.874, and 0.882 in terms of Mean Absolute Error (MAE) corresponding to C-cos, C-pcc and C-msd, respectively. The experimental results showed that their model outperforms several existing methods.

A recommender system (RS) is a subclass of information systems. It aims at providing the most relevant items (music, film…) that are preferred to each user. Several recommendation algorithms have been proposed in the literature and a comparison across their experimental results is necessary to evaluate the best algorithm. This paper presents a framework for presenting, developing and evaluating a recommender system. We preserve that this approach could play a vital role in elaborating an architecture and implementation of this type of systems. The proposed model presents the process of preparing the data set, whether rating or social data. It also includes a suite of state-of-the-art algorithms. The specificity of our architecture is the possibility of developing four kinds of recommender systems that are baseline, social, contextual and socio-contextual recommender system.

Nowadays, the Internet remains the essential source of access to scientific and technical information. In some languages, including Arabic, the means used to search for information do not seem to perform as well as in other languages. This deficiency is probably due to the late introduction of the Internet into the Arabization of the scientific and technological world, on the one hand, and to advances in the development of Arabic-language digital processing, on the other. This article intends to identify and explain the limitations and problems of searching for video in Arabic. For this reason, we propose an Arabic video indexing and retrieval system. In the indexation phase, our method is a combination between low-level and height-level indexation. In the retrieval phase, we put the user at the center of the research process. This goal is achieved by including a relevance feedback mechanism. The proposed new system is tested and prove that is perform well in a large scale database.

Indexing video by the concept is one of the most appropriate solutions for such problem. It's based on an association between a concept and its corresponding visual, sound or textual features. This kind of association is not a trivial task. It requires knowledge about the concept and its context. In this paper, we investigate a new concept detection approach to improve the performance of content-based multimedia documents retrieval systems. To achieve this goal, we tackle the problem from different plans and make four contributions at various stages of the indexing process. We first propose a new weakly supervised semi-automatic method based on the genetic algorithm to extract and annotate the video plans for training set. Subsequently, we develop a method to detect the basic concepts. We also deal with the issue of noise reduction when generating visual dictionary (BoVS). The different contributions are tested and evaluated on a big dataset (TRECVID 2015).

Following the technological advances carried out recently, there was an explosion in the number of available videos and their accessibility. This was largely justified by the fall of the acquisition prices of the memory supports and the increase their capacity, which facilitated the storage of a large number of video document databases in the computer systems. To make the exploitation of these collections effective, it is necessary to install tools that facilitate access and handle them. In this paper, we propose a new approach for semantic browsing in a large video collection. Our approach is based, essentially, on a structuring of the video concepts. Indeed, a weight is calculated to measure the membership degree of a concept at different levels (context / concept / video). We developed the Video Navigation System (VINAS) to test our approach. A visual metaphor inspired by tag cloud concept is also set up in our VINAS system to help with the comprehensive semantic video browsing. A user study made on our VINAS system showed that the proposed approach can help the user with the semantic browsing in video collection.

International audience

Recommender systems suggest the most appropriate items to users in order to help customers to find the most relevant items and facilitate sales. Collaborative filtering recommendation algorithm is the most successful technique for recommendation. In view of the fact that collaborative filtering systems depend on neighbors as the source of information, the recommendation quality of this approach depends on the neighbor’s selection. However, selecting neighbors can either stem from similarity or trust metrics. In this paper, we analyze these two types of neighbor’s selection metrics used in the field of recommendation in the literature. For each type, we first define it and then review different proposed metrics.

Nowadays, internet has offer an overabundance of available information. In social networks, users confront gigantic number of items. To overcome this phenomenon, known as information overload, recommender systems are intended to filter information and help users to make their choice. Many models based collaborative filtering have been used in the literature to solve the problem of recommendation. Among these models, latent factor model has become the most popular due to his performed results of accuracy. This work is part of research into Recommender System domain and aims to present a detailed explication on works based latent factor model. We first describe a general view of this model. Its realization in field of recommendation is next presented. A detailed study on different steps is then exposed. The most important works that have been developed are then presented. To the author’s knowledge, there has been no work that tries to explain in detail how latent factor model is applied to Recommender Systems.

International audience; Photo sharing in Mobile Ad-hoc Networks (MANETs) is becoming an important research topic in mobile computing due to (i) the progress of storage capacities of mobile devices such as cell phones and PDAs; and (ii) the intensive usage of cell phones to produce and share photos. Photo-sharing should be designed to work according to the connectivity lifetimes of nomadic users and their interests. As a photo-sharing in MANET is performed based on metadata, a photo sharing middleware should assist users during annotation of photos by taking users’ acquaintanceships into consideration. In this paper, a self-adaptive middleware called PASMi is proposed to carry out photo sharing and annotation activities by exploiting the interests of users, their connectivity-lifetimes and their acquaintanceships. The experimentations have been done to evaluate the performance of the PASMi middleware.

International audience; This paper proposes a new approach that uses social networks and common sense deduction rules to adapt the description tags of the photos for the current viewer. We exploit social graphs to enrich the tags associated to the concerned persons in the photo by following the different links between people (i.e. viewer and captured people in the photos). The main contributions of our work are: (i) addition of a more meaningful tagging layer for photos, making tags dynamic and auto-adaptable thanks to the automatic identification of the social context of the visualization. (ii) Due to this dynamics, the search in the social graphs is optimized using a data mining technique. (iii) we propose a new visualization metaphor for the tagging layer to manage users' feedback. We also describe a system architecture and an experimental study that shows significant improvements of the tagging process and execution times on a dataset containing triples in a FOAF graph.

International audience; PhotoMot is an online collaborative image management system which traces the actions of users and provides them assistance based on the capitalized experience according to the case based reasoning paradigm. How to imply the user in the assistance a system provides? PhotoMot shows the trace of user actions being built and lets users interact on the manner they are followed. This way they can help the system provide pertinent assistance. Manual image annotation and keyword based image search is a difficult and hardly formalizable problem. In this paper we present a model and a system which addresses this issue. The system provides basic online image gallery features: uploading single or zipped images, describing images with keywords, managing galleries … In addition the system helps image annotation tracing the actions of the user in order to propose additional keywords and images as well in the description as in the search phase. This help is provided by the capitalization and the reuse of users’ experience based on the case based reasoning paradigm. After a short introduction we describe our use trace model, the assistance strategies and present the developed prototype.

Concept detection has been considered a difficult problem and has attracted the interest of the content-based multimedia retrieval community. This detection implies an association between the concept and the visual content. In other words, the visual characteristics extracted from the video. This includes taking knowledge about the concept itself and its context. This work focuses on the problem of concept detection. For that, several stages are elaborated: first, a method of extraction and semi-automatic annotation of the video plans for the training set is proposed. This new method is based on the genetic algorithm. Then, a preliminary concept detection is carried out to generate the visual dictionary (BoVS). This second step is improved thanks to a noise reduction mechanism. This article's contribution has proven its effectiveness by testing it on a large dataset.