Rozprawy doktorskie na temat „Détection de contenu explicite”
Utwórz poprawne odniesienie w stylach APA, MLA, Chicago, Harvard i wielu innych
Sprawdź 22 najlepszych rozpraw doktorskich naukowych na temat „Détection de contenu explicite”.
Przycisk „Dodaj do bibliografii” jest dostępny obok każdej pracy w bibliografii. Użyj go – a my automatycznie utworzymy odniesienie bibliograficzne do wybranej pracy w stylu cytowania, którego potrzebujesz: APA, MLA, Harvard, Chicago, Vancouver itp.
Możesz również pobrać pełny tekst publikacji naukowej w formacie „.pdf” i przeczytać adnotację do pracy online, jeśli odpowiednie parametry są dostępne w metadanych.
Przeglądaj rozprawy doktorskie z różnych dziedzin i twórz odpowiednie bibliografie.
Vaglio, Andrea. "Leveraging lyrics from audio for MIR". Electronic Thesis or Diss., Institut polytechnique de Paris, 2021. http://www.theses.fr/2021IPPAT027.
Pełny tekst źródłaLyrics provide a lot of information about music since they encapsulate a lot of the semantics of songs. Such information could help users navigate easily through a large collection of songs and to recommend new music to them. However, this information is often unavailable in its textual form. To get around this problem, singing voice recognition systems could be used to obtain transcripts directly from the audio. These approaches are generally adapted from the speech recognition ones. Speech transcription is a decades-old domain that has lately seen significant advancements due to developments in machine learning techniques. When applied to the singing voice, however, these algorithms provide poor results. For a number of reasons, the process of lyrics transcription remains difficult. In this thesis, we investigate several scientifically and industrially difficult ’Music Information Retrieval’ problems by utilizing lyrics information generated straight from audio. The emphasis is on making approaches as relevant in real-world settings as possible. This entails testing them on vast and diverse datasets and investigating their scalability. To do so, a huge publicly available annotated lyrics dataset is used, and several state-of-the-art lyrics recognition algorithms are successfully adapted. We notably present, for the first time, a system that detects explicit content directly from audio. The first research on the creation of a multilingual lyrics-toaudio system are as well described. The lyrics-toaudio alignment task is further studied in two experiments quantifying the perception of audio and lyrics synchronization. A novel phonotactic method for language identification is also presented. Finally, we provide the first cover song detection algorithm that makes explicit use of lyrics information extracted from audio
Loukkas, Nassim. "Synthèse d'observateurs ensemblistes pour l’estimation d’état basées sur la caractérisation explicite des bornes d’erreur d’estimation". Thesis, Université Grenoble Alpes (ComUE), 2018. http://www.theses.fr/2018GREAT040/document.
Pełny tekst źródłaIn This work, we propose two main new approaches for the set-membershipstate estimation problem based on explicit characterization of the estimation error bounds. These approaches can be seen as a combination between a punctual observer and a setmembership characterization of the observation error. The objective is to reduce the complexity of the on-line implimentation, reduce the on-line computation time and improve the accuracy of the estimated state enclosure.The first approach is a set-membership observer based on ellipsoidal invariant sets for linear discrete-time systems and also for Linear Parameter Varying systems. The proposed approach provides a deterministic state interval that is build as the sum of the estimated system states and its corresponding estimation error bounds. The important feature of the proposed approach is that does not require propagation of sets.The second approach is an interval version of the Luenberger state observer for uncertain discrete-time linear systems based on interval and invariant set computation. The setmembership state estimation problem is considered as a punctual state estimation issue coupled with an interval characterization of the estimation error
Cámara, Chávez Guillermo. "Analyse du contenu vidéo par apprentissage actif". Cergy-Pontoise, 2007. http://www.theses.fr/2007CERG0380.
Pełny tekst źródłaThis thesis presents work towards a unified framework for semi-automated video indexing and interactive retrieval. To create an efficient index, a set of representative key frames are selected from the entire video content. We developed an automatic shot boundary detection algorithm to get rid of parameters and thresholds. We adopted a SVM classifier due to its ability to use very high dimensional feature spaces while at the same time keeping strong generalization guarantees from few training examples. We deeply evaluated the combination of features and kernels and present interesting results obtained, for shot extraction TRECVID 2006 Task. We then propose an interactive video retrieval system: RETINVID, to significantly reduce the number of key frames annotated by the user. The key frames are selected based on their ability to increase the knowledge of the data. We perform an experiment against the 2005 TRECVID benchmark for high-level task
Derbas, Nadia. "Contributions à la détection de concepts et d'événements dans les documents vidéos". Thesis, Grenoble, 2014. http://www.theses.fr/2014GRENM035/document.
Pełny tekst źródłaA consequence of the rise of digital technology is that the quantity of available collections of multimedia documents is permanently and strongly increasing. The indexing of these documents became both very costly and impossible to do manually. In order to be able to analyze, classify and search multimedia documents, indexing systems have been defined. However, most of these systems suffer quality or practicability issues. Their performance is limited and depends on the data volume and data variability. Indexing systems analyze multimedia documents, looking for static concepts (bicycle, chair,...), or events (wedding, protest,...). Therefore, the variability in shapes, positions, lighting or orientation of objects hinders the process. Another aspect is that systems must be scalable. They should be able to handle big data while using reasonable amount of computing time and memory.The aim of this thesis is to improve the general performance of content-based multimedia indexing systems. Four main contributions are brought in this thesis for improving different stages of the indexing process. The first one is an "early-early fusion method" that merges different information sources in order to extract their deep correlations. This method is used for violent scenes detection in movies. The second contribution is a weakly supervised method for basic concept (objects) localization in images. This can be used afterwards as a new descriptor to help detecting complex concepts (events). The third contribution tackles the noise reduction problem on ambiguously annotated data. Two methods are proposed: a shot annotation generator, and a shot weighing method. The last contribution is a generic descriptor optimization method, based on PCA and non-linear transforms.These four contributions are tested and evaluated using reference data collections, including TRECVid and MediaEval. These contributions helped our submissions achieving very good rankings in those evaluation campaigns
Carlier, Axel. "Compréhension de contenus visuels par analyse conjointe du contenu et des usages". Thesis, Toulouse, INPT, 2014. http://www.theses.fr/2014INPT0085/document.
Pełny tekst źródłaThis thesis focuses on the problem of understanding visual contents, which can be images, videos or 3D contents. Understanding means that we aim at inferring semantic information about the visual content. The goal of our work is to study methods that combine two types of approaches: 1) automatic content analysis and 2) an analysis of how humans interact with the content (in other words, usage analysis). We start by reviewing the state of the art from both Computer Vision and Multimedia communities. Twenty years ago, the main approach was aiming at a fully automatic understanding of images. This approach today gives way to different forms of human intervention, whether it is through the constitution of annotated datasets, or by solving problems interactively (e.g. detection or segmentation), or by the implicit collection of information gathered from content usages. These different types of human intervention are at the heart of modern research questions: how to motivate human contributors? How to design interactive scenarii that will generate interactions that contribute to content understanding? How to check or ensure the quality of human contributions? How to aggregate human contributions? How to fuse inputs obtained from usage analysis with traditional outputs from content analysis? Our literature review addresses these questions and allows us to position the contributions of this thesis. In our first set of contributions we revisit the detection of important (or salient) regions through implicit feedback from users that either consume or produce visual contents. In 2D, we develop several interfaces of interactive video (e.g. zoomable video) in order to coordinate content analysis and usage analysis. We also generalize these results to 3D by introducing a new detector of salient regions that builds upon simultaneous video recordings of the same public artistic performance (dance show, chant, etc.) by multiple users. The second contribution of our work aims at a semantic understanding of fixed images. With this goal in mind, we use data gathered through a game, Ask’nSeek, that we created. Elementary interactions (such as clicks) together with textual input data from players are, as before, mixed with automatic analysis of images. In particular, we show the usefulness of interactions that help revealing spatial relations between different objects in a scene. After studying the problem of detecting objects on a scene, we also adress the more ambitious problem of segmentation
Hamroun, Mohamed. "Indexation et recherche par contenu visuel, sémantique et multi-niveaux des documents multimédia". Thesis, Bordeaux, 2019. http://www.theses.fr/2019BORD0372.
Pełny tekst źródłaDue to the latest technological advances, the amount of multimedia data is constantly increasing. In this context, the problem is how to effectively use this data? it is necessary to set up tools to facilitate its access and manipulation.To achieve this goal, we first propose an indexation and retrieval model for video shots (or images) by their visual content (ISE). The innovative features of ISE are as follows: (i) definition of a new descriptor "PMC" and (ii) application of the genetic algorithm (GA) to improve the retrieval (PMGA).Then, we focus on the detection of concepts in video shots (LAMIRA approach). In the same context, we propose a semi-automatic annotation method for video shots in order to improve the quality of indexation based on the GA.Then, we provide a semantic indexation method separating the data level from a conceptual level and a more abstract, contextual level. This new system also incorporates mechanisms for expanding the request and relevance feedback. To add more fluidity to the user query, the user can perform a navigation using the three levels of abstraction. Two systems called VISEN and VINAS have been set up to validate these last positions.Finally, a SIRI Framework was proposed on the basis of a multi-level indexation combining our 3 systems: ISE, VINAS and VISEN. This Framework provides a two-dimensional representation of features (high level and low level) for each image
Artaud, Chloé. "Détection des fraudes : de l’image à la sémantique du contenu : application à la vérification des informations extraites d’un corpus de tickets de caisse". Thesis, La Rochelle, 2019. http://www.theses.fr/2019LAROS002/document.
Pełny tekst źródłaCompanies, administrations, and sometimes individuals, have to face many frauds on documents they receive from outside or process internally. Invoices, expense reports, receipts...any document used as proof can be falsified in order to earn more money or not to lose it. In France, losses due to fraud are estimated at several billion euros per year. Since the flow of documents exchanged, whether digital or paper, is very important, it would be extremely costly and time-consuming to have them all checked by fraud detection experts. That’s why we propose in our thesis a system for automatic detection of false documents. While most of the work in automatic document detection focuses on graphic clues, we seek to verify the textual information in the document in order to detect inconsistencies or implausibilities.To do this, we first compiled a corpus of documents that we digitized. After correcting the characters recognition outputs and falsifying part of the documents, we extracted the information and modelled them in an ontology, in order to keep the semantic links between them. The information thus extracted, and increased by its possible disambiguation, can be verified against each other within the document and through the knowledge base established. The semantic links of ontology also make it possible to search for information in other sources of knowledge, particularly on the Internet
Sahli, Samir. "Extraction de l'objet de référence par la transformation multiéchelle Beamlet : détection de pistes d'atterrissage dans une image aérienne". Thesis, Université Laval, 2008. http://www.theses.ulaval.ca/2008/25100/25100.pdf.
Pełny tekst źródłaMeng, Zide. "Analyse temporelle et sémantique des réseaux sociaux typés à partir du contenu de sites généré par des utilisateurs sur le Web". Thesis, Université Côte d'Azur (ComUE), 2016. http://www.theses.fr/2016AZUR4090/document.
Pełny tekst źródłaWe propose an approach to detect topics, overlapping communities of interest, expertise, trends andactivities in user-generated content sites and in particular in question-answering forums such asStackOverFlow. We first describe QASM (Question & Answer Social Media), a system based on socialnetwork analysis to manage the two main resources in question-answering sites: users and contents. Wealso introduce the QASM vocabulary used to formalize both the level of interest and the expertise ofusers on topics. We then propose an efficient approach to detect communities of interest. It relies onanother method to enrich questions with a more general tag when needed. We compared threedetection methods on a dataset extracted from the popular Q&A site StackOverflow. Our method basedon topic modeling and user membership assignment is shown to be much simpler and faster whilepreserving the quality of the detection. We then propose an additional method to automatically generatea label for a detected topic by analyzing the meaning and links of its bag of words. We conduct a userstudy to compare different algorithms to choose the label. Finally we extend our probabilistic graphicalmodel to jointly model topics, expertise, activities and trends. We performed experiments with realworlddata to confirm the effectiveness of our joint model, studying the users’ behaviors and topicsdynamics
Ollagnier, Anaïs. "Analyse de requêtes en langue naturelle et extraction d'informations bibliographiques pour une recherche de livres orientée contenu efficace". Thesis, Aix-Marseille, 2017. http://www.theses.fr/2017AIXM0556/document.
Pełny tekst źródłaIn the recent years, the Web has undergone a tremendous growth regarding both content and users. This has led to an information overload problem in which people are finding it increasingly difficult to locate the right information at the right time. Recommender systems have been developed to address this problem, by guiding users through the big ocean of information. The recommendation approaches have multiplied and have been successfully implemented, particularly through approaches such as collaborative filtering. However, there are still challenges and limitations that offer opportunities for new research. Among these challenges, the design of reading recommendation systems has become a new expanding research focus following the emergence of digital libraries.Traditionally, libraries play a passive role in interaction with users due to the lack of effective search and recommendation tools. In this manuscript, we will study the creation of a reading recommendation system in which we'll try to exploit the possibilities of digital access to scientific information. Our objectives are: - to improve the understanding of user needs expressed in natural language search queries for books, articles and posts. This work will require the establishment of processes capable of exploiting the structures of data and their dimension; - to compensate for the absence of explicit links between books and journal articles by automatically detecting and analyzing bibliographic references, and then to propose links;- to achieve a reading recommendation system based on textual data to provide a customized recommendation list to active users, similar to systems already used by users profiles
Vicente, David. "Modèles de Mumford-Shah pour la détection de structures fines en image". Thesis, Orléans, 2015. http://www.theses.fr/2015ORLE2055/document.
Pełny tekst źródłaThis thesis is a contribution to the fine tubular structures detection problem in a 2-D or 3-D image. We arespecifically interested in the case of angiographic images. The vessels are surrounded by noise and thenthe question is to segment precisely the blood network. The theoretical framework of our work is thecalculus of variations and we focus on the Mumford-Shah energy. Initially, this model is adapted to thedetection of volumetric structures extended in all directions of the image. The aim of this study is to buildan energy that favors sets which are extended in one direction, which is the case of fine tubes. Then, weintroduce a new unknown, a Riemannian metric, which captures the geometric structure at each point ofthe image and we give a new formulation of the Mumford-Shah energy adapted to this metric. Thecomplete analysis of this model is done: we prove that the associated problem of minimization is wellposed and we introduce an approximation by gamma-convergence more suitable for numerics. Eventually,we propose numerical experimentations adapted to this approximation
Niaz, Usman. "Amélioration de la détection des concepts dans les vidéos en coupant de plus grandes tranches du monde visuel". Thesis, Paris, ENST, 2014. http://www.theses.fr/2014ENST0040/document.
Pełny tekst źródłaVisual material comprising images and videos is growing ever so rapidly over the internet and in our personal collections. This necessitates automatic understanding of the visual content which calls for the conception of intelligent methods to correctly index, search and retrieve images and videos. This thesis aims at improving the automatic detection of concepts in the internet videos by exploring all the available information and putting the most beneficial out of it to good use. Our contributions address various levels of the concept detection framework and can be divided into three main parts. The first part improves the Bag of Words (BOW) video representation model by proposing a novel BOW construction mechanism using concept labels and by including a refinement to the BOW signature based on the distribution of its elements. We then devise methods to incorporate knowledge from similar and dissimilar entities to build improved recognition models in the second part. Here we look at the potential information that the concepts share and build models for meta-concepts from which concept specific results are derived. This improves recognition for concepts lacking labeled examples. Lastly we contrive certain semi-supervised learning methods to get the best of the substantial amount of unlabeled data. We propose techniques to improve the semi-supervised cotraining algorithm with optimal view selection
Niaz, Usman. "Amélioration de la détection des concepts dans les vidéos en coupant de plus grandes tranches du monde visuel". Electronic Thesis or Diss., Paris, ENST, 2014. http://www.theses.fr/2014ENST0040.
Pełny tekst źródłaVisual material comprising images and videos is growing ever so rapidly over the internet and in our personal collections. This necessitates automatic understanding of the visual content which calls for the conception of intelligent methods to correctly index, search and retrieve images and videos. This thesis aims at improving the automatic detection of concepts in the internet videos by exploring all the available information and putting the most beneficial out of it to good use. Our contributions address various levels of the concept detection framework and can be divided into three main parts. The first part improves the Bag of Words (BOW) video representation model by proposing a novel BOW construction mechanism using concept labels and by including a refinement to the BOW signature based on the distribution of its elements. We then devise methods to incorporate knowledge from similar and dissimilar entities to build improved recognition models in the second part. Here we look at the potential information that the concepts share and build models for meta-concepts from which concept specific results are derived. This improves recognition for concepts lacking labeled examples. Lastly we contrive certain semi-supervised learning methods to get the best of the substantial amount of unlabeled data. We propose techniques to improve the semi-supervised cotraining algorithm with optimal view selection
Bendraou, Youssef. "Détection des changements de plans et extraction d'images représentatives dans une séquence vidéo". Thesis, Littoral, 2017. http://www.theses.fr/2017DUNK0458/document.
Pełny tekst źródłaWith the recent advancement in multimedia technologies, in conjunction with the rapid increase of the volume of digital video data and the growth of internet ; it has becom mandatory to have the hability browse and search through information stored in large multimedia databases. For this purpose, content based video retrieval (CBVR) has become an active area of research durinf the last decade. The objective of this thesis is to present applications for temporal video segmentation and video retrieval based on different mathematical models. A shot is considered as the elementary unit of a video, and is defined as a continuous sequence of frames taken from a single camera, representing an action during time. The different types of transitions that may occur in a video sequence are categorized into : abrupt and gradual transition. In this work, through statistical analysis, we segment a video into its constituent units. This is achieved by identifying transitions between adjacent shots. The first proposed algorithm aims to detect abrupt shot transitions only by measuring the similarity between consecutive frames. Given the size of the vector containing distances, it can be modeled by a log normal distribution since all the values are positive. Gradual shot transition identification is a more difficult task when compared to cut detection. Generally, a gradual transition may share similar characteristics as a dynamic segment with camera or object motion. In this work, singular value decomposition (SVD) is performed to project features from the spatial domain to the singular space. Resulting features are reduced and more refined, which makes the remaining tasks easier. The proposed system, designed for detecting both abrupt and gradual transitions, has lead to reliable performances achieving high detection rates. In addition, the acceptable computational time allows to process in real time. Once a video is partitioned into its elementary units, high-level applications can be processed, such as the key-frame extraction. Selecting representative frames from each shot to form a storyboard is considered as a static and local video summarization. In our research, we opted for a global method based on local extraction. Using refined centrist features from the singular space, we select representative frames using modified k-means clustering based on important scenes. This leads to catch pertinent frames without redoudancy in the final storyboard
Kassab, Randa. "Analyse des propriétés stationnaires et des propriétés émergentes dans les flux d'information changeant au cours du temps". Thesis, Nancy 1, 2009. http://www.theses.fr/2009NAN10027/document.
Pełny tekst źródłaMany applications produce and receive continuous, unlimited, and high-speed data streams. This raises obvious problems of storage, treatment and analysis of data, which are only just beginning to be treated in the domain of data streams. On the one hand, it is a question of treating data streams on the fly without having to memorize all the data. On the other hand, it is also a question of analyzing, in a simultaneous and concurrent manner, the regularities inherent in the data stream as well as the novelties, exceptions, or changes occurring in this stream over time. The main contribution of this thesis concerns the development of a new machine learning approach - called ILoNDF - which is based on novelty detection principle. The learning of this model is, contrary to that of its former self, driven not only by the novelty part in the input data but also by the data itself. Thereby, ILoNDF can continuously extract new knowledge relating to the relative frequencies of the data and their variables. This makes it more robust against noise. Being operated in an on-line mode without repeated training, ILoNDF can further address the primary challenges for managing data streams. Firstly, we focus on the study of ILoNDF's behavior for one-class classification when dealing with high-dimensional noisy data. This study enabled us to highlight the pure learning capacities of ILoNDF with respect to the key classification methods suggested until now. Next, we are particularly involved in the adaptation of ILoNDF to the specific context of information filtering. Our goal is to set up user-oriented filtering strategies rather than system-oriented in following two types of directions. The first direction concerns user modeling relying on the model ILoNDF. This provides a new way of looking at user's need in terms of specificity, exhaustivity and contradictory profile-contributing criteria. These criteria go on to estimate the relative importance the user might attach to precision and recall. The filtering threshold can then be adjusted taking into account this knowledge about user's need. The second direction, complementary to the first one, concerns the refinement of ILoNDF's functionality in order to confer it the capacity of tracking drifting user's need over time. Finally, we consider the generalization of our previous work to the case where streaming data can be divided into multiple classes
Guironnet, Mickaël. "Méthodes de résumé de vidéo à partir d'informations bas niveau, du mouvement de caméra ou de l'attention visuelle". Université Joseph Fourier (Grenoble), 2006. http://www.theses.fr/2006GRE10155.
Pełny tekst źródłaThe growing volume of video leads to the need of new tools for indexing. One of the possible tools is video summary which provides a fast overview to the user. The objective of this thesis is to extract from visual information, a summary containing the “message” of video. We chose to study three new methods of video summary using different types of visual features. The first method of summary rests on low level features (color, orientation and motion). The combination of these features which is based on a fuzzy inference system allows a hierarchical summary to be built. We show the interest of such a summary in an application of query by example. The second method of summary is built from camera motion. This higher level feature is thought by the filmmaker and so induces information on the content. A method of camera motion classification based on Transferable Belief Model is achieved. The method of summary is elaborated according to rules about the magnitude and the chain of the identified motions. The third method of summary is developed from visual attention. To know the places where the glance is directed during the video playback is higher level information and relevant to create the summary. A spatio-temporal attention model is proposed, and then used to detect the change of content in time in order to build the summary
Alves, do Valle Junior Eduardo. "Local-Descriptor Matching for Image Identification Systems". Cergy-Pontoise, 2008. http://biblioweb.u-cergy.fr/theses/08CERG0351.pdf.
Pełny tekst źródłaImage identification (or copy detection) consists in retrieving the original from which a query image possibly derives, as well as any related metadata, such as titles, authors, copyright information, etc. The task is challenging because of the variety of transformations that the original image may have suffered. Image identification systems based on local descriptors have shown excellent efficacy, but often suffer from efficiency issues, since hundreds, even thousands of descriptors, have to be matched in order to find a single image. The objective of our work is to provide fast methods for descriptor matching, by creating efficient ways to perform the k-nearest neighbours search in high-dimensional spaces. In this way, we can gain the advantages from the use of local descriptors, while minimising the efficiency issues. We propose three new methods for the k-nearest neighbours search: the 3-way trees — an improvement over the KD-trees using redundant, overlapping nodes; the projection KD-forests — a technique which uses multiple moderate dimensional KD-trees; and the multicurves, which is based on multiple moderate dimensional Hilbert space-filling curves. Those techniques try to reduce the amount of random access to the data, in order to be well adapted to the implementation in secondary memory
Firoozeh, Nazanin. "Semantic-oriented Recommandation for Content Enrichment". Thesis, Sorbonne Paris Cité, 2018. http://www.theses.fr/2018USPCD033.
Pełny tekst źródłaIn this thesis, we aim at enriching the content of an unstructured document with respect to a domain of interest. The goal is to minimize the vocabulary and informational gap between the document and the domain. Such an enrichment which is based on Natural Language Processing and Information Retrieval technologies has several applications. As an example, flling in the gap between a scientifc paper and a collection of highly cited papers in a domain helps the paper to be better acknowledged by the community that refers to that collection. Another example is to fll in the gap between a web page and the usual keywords of visitors that are interested in a given domain so as it is better indexed and referred to in that domain, i.e. more accessible for those visitors. We propose a method to fll that gap. We first generate an enrichment collection, which consists of the important documents related to the domain of interest. The main information of the enrichment collection is then extracted, disambiguated and proposed to a user,who performs the enrichment. This is achieved by decomposing the problem into two main components of keyword extraction and topic detection. We present a comprehensive study over different approaches of each component. Using our findings, we propose approaches for extracting keywords from web pages, detecting their under lying topics, disambiguating them and returning the ones related to the domain of interest. The enrichment is performed by recommending discriminative sets of semantically relevant keywords, i.e. topics, to a user. The topics are labeled with representative keywords and have a level of granularity that is easily interpretable. Topic keywords are ranked by importance. This helps to control the length of the document, which needs to be enriched, by targeting the most important keywords of each topic. Our approach is robust to the noise in web pages. It is also knowledge-poor and domain-independent. It, however, exploits search engines for generating the required data but is optimized in the number of requests sent to them. In addition, the approach is easily tunable to different languages. We have implemented the keyword extraction approach in 12 languages and four of them have been tested over various domains. The topic detection approach has been implemented and tested on English and French. However, it is on French language that the approaches have been tested on a large scale : the keyword extraction on roughly 400 domains and the topic detection on 80 domains.To evaluate the performance of our enrichment approach, we focused on French and we performed different experiments on the proposed keyword extraction and topic detection methods. To evaluate their robustness, we studied them on 10 topically diverse domains.Results were evaluated through both user-based evaluations on a real application context and by comparing with baseline approaches. Our results on the keyword extraction approach showed that the statistical features are not adequate for capturing words importance within a web page. In addition, we found our proposed approach of keyword extraction to be effective when applied on real applications. The evaluations on the topic detection approach also showed that it can electively filter out the keywords which are not related to a target domain and that it labels the topics with representative and discriminative keywords. In addition, the approach achieved a high precision in preserving the semantic consistency of the keywords within each topic. We showed that our approach out performs a baseline approach, since the widely-used co-occurrence feature between keywords is notivenough for capturing their semantic similarity and consequently for detecting semantically consistent topics
Kassab, Randa. "Analyse des propriétés stationnaires et des propriétés émergentes dans les flux d'informations changeant au cours du temps". Phd thesis, Université Henri Poincaré - Nancy I, 2009. http://tel.archives-ouvertes.fr/tel-00402644.
Pełny tekst źródłaL'apport de ce travail de thèse réside principalement dans le développement d'un modèle d'apprentissage - nommé ILoNDF - fondé sur le principe de la détection de nouveauté. L'apprentissage de ce modèle est, contrairement à sa version de départ, guidé non seulement par la nouveauté qu'apporte une donnée d'entrée mais également par la donnée elle-même. De ce fait, le modèle ILoNDF peut acquérir constamment de nouvelles connaissances relatives aux fréquences d'occurrence des données et de leurs variables, ce qui le rend moins sensible au bruit. De plus, doté d'un fonctionnement en ligne sans répétition d'apprentissage, ce modèle répond aux exigences les plus fortes liées au traitement des flux de données.
Dans un premier temps, notre travail se focalise sur l'étude du comportement du modèle ILoNDF dans le cadre général de la classification à partir d'une seule classe en partant de l'exploitation des données fortement multidimensionnelles et bruitées. Ce type d'étude nous a permis de mettre en évidence les capacités d'apprentissage pures du modèle ILoNDF vis-à-vis de l'ensemble des méthodes proposées jusqu'à présent. Dans un deuxième temps, nous nous intéressons plus particulièrement à l'adaptation fine du modèle au cadre précis du filtrage d'informations. Notre objectif est de mettre en place une stratégie de filtrage orientée-utilisateur plutôt qu'orientée-système, et ceci notamment en suivant deux types de directions. La première direction concerne la modélisation utilisateur à l'aide du modèle ILoNDF. Cette modélisation fournit une nouvelle manière de regarder le profil utilisateur en termes de critères de spécificité, d'exhaustivité et de contradiction. Ceci permet, entre autres, d'optimiser le seuil de filtrage en tenant compte de l'importance que pourrait donner l'utilisateur à la précision et au rappel. La seconde direction, complémentaire de la première, concerne le raffinement des fonctionnalités du modèle ILoNDF en le dotant d'une capacité à s'adapter à la dérive du besoin de l'utilisateur au cours du temps. Enfin, nous nous attachons à la généralisation de notre travail antérieur au cas où les données arrivant en flux peuvent être réparties en classes multiples.
Tirilly, Pierre. "Traitement automatique des langues pour l'indexation d'images". Phd thesis, Université Rennes 1, 2010. http://tel.archives-ouvertes.fr/tel-00516422.
Pełny tekst źródłaTirilly, Pierre. "Traitement automatique des langues pour l'indexation d'images". Phd thesis, Rennes 1, 2010. http://www.theses.fr/2010REN1S045.
Pełny tekst źródłaIn this thesis, we propose to integrate natural language processing (NLP) techniques in image indexing systems. We first address the issue of describing the visual content of images. We rely on the visual word-based image description, which raises problems that are well known in the text indexing field. First, we study various NLP methods (weighting schemes and stop-lists) to automatically determine which visual words are relevant to describe the images. Then we use language models to take account of some geometrical relations between the visual words. We also address the issue of describing the semantic content of images: we propose an image annotation scheme that relies on extracting relevant named entities from texts coming with the images to annotate
Nguyen, Thanh-Khoa. "Image segmentation and extraction based on pixel communities". Thesis, La Rochelle, 2019. http://www.theses.fr/2019LAROS035.
Pełny tekst źródłaImage segmentation has become an indispensable task that is widely employed in several image processing applications including object detection, object tracking, automatic driver assistance, and traffic control systems, etc. The literature abounds with algorithms for achieving image segmentation tasks. These methods can be divided into some main groups according to the underlying approaches, such as Region-based image segmentation, Feature-based clustering, Graph-based approaches and Artificial Neural Network-based image segmentation. Recently, complex networks have mushroomed both theories and applications as a trend of developments. Hence, image segmentation techniques based on community detection algorithms have been proposed and have become an interesting discipline in the literature. In this thesis, we propose a novel framework for community detection based image segmentation. The idea that brings social networks analysis domain into image segmentation quite satisfies with most authors and harmony in those researches. However, how community detection algorithms can be applied in image segmentation efficiently is a topic that has challenged researchers for decades. The contribution of this thesis is an effort to construct best complex networks for applying community detection and proposal novel agglomerate methods in order to aggregate homogeneous regions producing good image segmentation results. Besides, we also propose a content based image retrieval system using the same features than the ones obtained by the image segmentation processes. The proposed image search engine for real images can implement to search the closest similarity images with query image. This content based image retrieval relies on the incorporation of our extracted features into Bag-of-Visual-Words model. This is one of representative applications denoted that image segmentation benefits several image processing and computer visions applications. Our methods have been tested on several data sets and evaluated by many well-known segmentation evaluation metrics. The proposed methods produce efficient image segmentation results compared to the state of the art