Développer des ressources lexico-sémantiques pour le Traitement Automatique des Langues Naturelles est un enjeu majeur du domaine. Ces ressources explicitant notamment des connaissances que seuls les humains possèdent, ont pour but de permettre aux applications de TALNune compréhension de texte assez fine et complète. De nouvelles approches populaires de construction de ces dernières impliquant l'externalisation ouverte (crowdsourcing) émergent en TALN. Elles ont confirmé leur efficacité et leur pertinence. Cependant, les ressources obtenues ne sont pas exemptes d'informations erronées ou de silences causés par l'absence de certaines relations sémantiques pertinentes et primordiales pour la bonne qualité. Dans ce travail de recherche, nous prenons comme exemple d'étude le réseau lexico-sémantique du projet JeuxDeMots et nous proposons un système de consolidation endogène pour ce type de réseaux.Ce système se base principalement sur l'enrichissement du réseau par l'inférence et l'annotation de nouvelles relations à partir de celles existantes, ainsi que l'extraction de règles d'inférence permettant de (re)générer une grande partie du réseau. Enfin, un langage dédié de manipulation du système de consolidation et du réseau lexico-sémantique est conçu et un premier prototype a été implémenté. / Developing lexico-semantic resources is a major issue in the Natural Language Processing field.These resources, by making explicit inter alia some knowledge possessed only by humans, aim at providing the ability of a precise and complete text understanding to NLP tasks. Popular resources-building strategies involving crowdsourcing are flowering in NLP and are proved to be successful. However, the resulted resources are not free of errors and lack some important semantic relations. In this PhD thesis, we used the french lexico-semantic network from the project JeuxDeMots as a case-study. We designed an endogenous consolidation system for this type of networks based on inferring and annotating new semantic relations using the already existing ones, as well as extracting and proposing inference rules able to (re)generate a considerable part of the network. In addition, we conceived a domain specific language for manipulating the consolidation system along with the network itself and a prototype was implemented.

Made available in DSpace on 2015-03-05T14:01:20Z (GMT). No. of bitstreams: 0 Previous issue date: 19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Composição de serviços em dispositivos móveis está se tornando indispensável para as atuais e futuras necessidades dos usuários de TI. Estas necessidades são suportadas pela convergência de arquiteturas centralizadas para arquiteturas distribuídas que estimulam a composição de serviços web. Estes serviços web, motivados pela arquitetura Web Services, logo evoluíram para SOA. Esta arquitetura por sua vez, no contexto de orquestração de serviços, tem exigido linguagens de composição para sua efetiva implementação. Linguagens estas, que têm progredido para linguagens interpretadas e declarativas. Neste contexto de evolução, este trabalho apresenta, uma linguagem declarativa de composição de serviços chamada inSOA. Esta linguagem foi desenvolvida para ser executada em dispositivos móveis como parte de um projeto de uma arquitetura ubíqua colaborativa chamada de U-SOA. inSOA é uma solução focada em Web Services com vocação SOA e aderente aos patterns de composição de serviços, além de poder ser embutida em lin / Services composition on mobile devices is becoming essential for current and future needs of IT users. These needs are supported by the convergence from centralized to distributed architectures that encourage the web services composition. These web services, motivated by the Web Services architecture, soon evolved into an SOA. This architecture in turn, in the services orchestration, has required composition language for their effective implementation. These languages, has advanced to interpreted and declarative languages. In this evolution context, this work presents, a declarative language of services composition services called inSOA. This language was developed to execute on mobile devices as part of a project of a collaborative ubiquitous architecture called USOA. InSOA is a solution focused on Web Services and SOA adherent to the patterns of services composition, and can be embedded in a language of general purpose and light in the better DSL languages style. As a way to test and evaluate the langu

Dans cette thèse nous nous sommes intéressés aux problèmes concernant la conception et la gestion des plans de soins à domicile. Un plan de soins à domicile définit l'ensemble des activités médicales et/ou sociales qui sont menées jour après jour au domicile d'un patient. Ce plan de soins est généralement construit à travers un processus complexe impliquant une évaluation complète des besoins du patient ainsi que son environnement social et physique. La spécification de plans de soins à domicile est difficile pour plusieurs raisons: les plans de soins à domicile sont par nature des processus non structurés qui impliquent des activités répétitives mais irrégulières, dont la spécification requiert des expressions temporelles complexes. Ces caractéristiques font que les plans de soins à domicile sont difficiles à modéliser en utilisant les technologies traditionnelles de modélisation de processus. Tout d'abord, nous présentons l'approche basée sur les DSL (Langage spécifique au domaine) qui permet d'exprimer les plans de soins à domicile en utilisant des abstractions de haut niveau et orientées utilisateur. Le DSL nous permet à travers cette thèse de proposer un langage de temporalités permettant de spécifier les temporalités des activités du plan de soins à domicile. Ensuite, nous décrivons comment les plans de soins à domicile, formalisés grâce aux automates temporisés, peuvent être générés à partir de ces abstractions. Nous proposons une approche en trois étapes qui consiste à: (i) le mapping entre les spécifications temporelles élémentaires et les automates temporisés appelés "pattern automata", (ii) la combinaison des "patterns automata" afin de construire les automates d'activités en utilisant l'algorithme de composition que nous avons déni, et aussi (iii) la construction de l'automate de plan de soins à domicile global. L'automate de plan de soins à domicile résultant englobe tous les schedules autorisés des activités pour un patient donné. Enfin, nous montrons comment la vérification et le suivi de l'automate du plan de soins à domicile résultant peuvent être faits en utilisant des techniques et des outils existants, en particulier en utilisant l'outil de verification UPPAAL. / In this thesis we are interested in the problems underlying the design and the management of home care plans. A home care plan defines the set of medical and/or social activities that are carried out day after day at a patient's home. Such a care plan is usually constructed through a complex process involving a comprehensive assessment of patient's needs as well as his/her social and physical environment. Specication of home care plans is challenging for several reasons: home care plans are inherently nonstructured processes which involve repetitive, but irregular, activities, whose specification requires complex temporal expressions. These features make home care plans difficult to model using traditional process modeling technologies. First, we present a DSL (Domain Specific Language) based approach tailored to express home care plans using high level and user-oriented abstractions. DSL enables us through this thesis to propose a temporalities language to specify temporalities of home care plan activities. Then, we describe how home care plans, formalized as timed automata, can be generated from these abstractions. We propose a three-step approach which consists in (i) mapping between elementary temporal specifications and timed automata called Pattern automata, (ii) combining patterns automata to build the activity automata using our composition algorithm and then (iii) constructing the global care plan automaton. The resulting care plan automaton encompasses all the possible allowed schedules of activities for a given patient. Finally, we show how verification and monitoring of the resulting care plan can be handled using existing techniques and tools, especially using UPPAAL Model Checker.

領域專屬語言的設計理念是希望能夠協助使用者在特定領域上解決特定問題。然而，大多數領域專屬語言的開發環境與工具均非常貧乏，這將會增加使用者在開發程式上的困難。 所以本研究利用模型導向技術來建構一套生成系統，使用者只要輸入領域專屬語言的語意以及偵錯定義，就能快速生成語言的執行與偵錯環境，並提供完善的操作介面，輔助使用者加快程式開發的速度。 / The purpose of creating domain-specific languages (DSLs) is to help user solve problems in a particular domain. However, most DSLs are lack of development environments and tools, which would be more difficult to develop programs. This thesis is aimed to develop a generating system based on model-driven technology. Given the semantics and debugging definitions of a domain-specific language, the system would be able to generate an executing and debugging environment for the language with friendly user interface, thus improving efficiency and productivity of using the language.

The major problem addressed in this thesis is the definition of a new SPLE methodology that covers both phases of the SPLE process and focuses on the derivation of behavioral models of SPL products. In Chapter 2 three research areas scope context of this thesis: Software Product Lines, Business Processes, and Model-Driven Engineering. Throughout Chapter 3, we propose a new SPLE methodology that focuses on the derivation of product behavior. We first describe the main flow of the methodology, and then detail the individual steps. In chapter 4 we propose a new domain specific language called CBPF created for modeling composable business process fragments. A model driven approach is followed for creating CBPF: definition of the abstract syntax, graphical concrete syntax and translational semantics. In Chapter 5 we propose several types of verifications that can be applied to business processfragments to determine their "correctness". For structural verification we definine a set of fragment consistency rules that should be valid for every business process fragment created with CBPF. To check behavioral correctness we first transform the business process fragment into an equivalent HCPN. We can then check generic properties but also define aset of fragment specific properties. In chapter 6 we exemplify the proposed SPL methodology by applying it to a case study from the crisis management system domain. We also propose a tool suite that supports our methodology. Chapter 7 describes possible improvements and extensions to the contributions of this thesis. We conclude the thesis in Chapter 8 and draw some conclusions.

Geometric Multigrid (GMG) methods are widely used in numerical analysis to accelerate the convergence of partial differential equations solvers using a hierarchy of grid discretizations. These solvers find plenty of applications in various fields in engineering and scientific domains, where solving PDEs is of fundamental importance. Using multigrid methods, the pace at which the solvers arrive at the solution can be improved at an algorithmic level. With the advance in modern computer architecture, solving problems with higher complexity and sizes is feasible - this is also the case with multigrid methods. However, since hardware support alone cannot achieve high performance in execution time, there is a need for good software that help programmers in doing so. Multiple grid sizes and recursive expression of multigrid cycles make the task of manual program optimization tedious and error-prone. A high-level language that aids domain experts to quickly express complex algorithms in a compact way using dedicated constructs for multigrid methods and with good optimization support is thus valuable. Typical computation patterns in a GMG algorithm includes stencils, point-wise accesses, restriction and interpolation of a grid. These computations can be optimized for performance on modern architectures using standard parallelization and locality enhancement techniques. Several past works have addressed the problem of automatic optimizations of computations in various scientific domains using a domain-specific language (DSL) approach. A DSL is a language with features to express domain-specific computations and compiler support to enable optimizations specific to these computations. Halide and PolyMage are two of the recent works in this direction, that aim to optimize image processing pipelines. Many computations like upsampling and downsampling an image are similar to interpolation and restriction in geometric multigrid methods. In this thesis, we demonstrate how high performance can be achieved on GMG algorithms written in the PolyMage domain-specific language with new optimizations we added to the compiler. We also discuss the implementation of non-trivial optimizations, on PolyMage compiler, necessary to achieve high parallel performance for multigrid methods on modern architectures. We realize these goals by: • introducing multigrid domain-specific constructs to minimize the verbosity of the algorithm specification; • storage remapping to reduce the memory footprint of the program and improve cache locality exploitation; • mitigating execution time spent in data handling operations like memory allocation and freeing, using a pool of memory, across multiple multigrid cycles; and • incorporating other well-known techniques to leverage performance, like exploiting multi-dimensional parallelism and minimizing the lifetime of storage buffers. We evaluate our optimizations on a modern multicore system using five different benchmarks varying in multigrid cycle structure, complexity and size, for two-and three-dimensional data grids. Experimental results show that our optimizations: • improve performance of existing PolyMage optimizer by 1.31x; • are better than straight-forward parallel and vector implementations by 3.2x; • are better than hand-optimized versions in conjunction with optimizations by Pluto, a state-of-the-art polyhedral source-to-source optimizer, by 1.23x; and • achieve up to 1.5$\times$ speedup over NAS MG benchmark from the NAS Parallel Benchmarks. (The speedup numbers are Geometric means over all benchmarks)

Graphs model relationships across real-world entities in web graphs, social network graphs, and road network graphs. Graph algorithms analyze and transform a graph to discover graph properties or to apply a computation. For instance, a pagerank algorithm computes a rank for each page in a webgraph, and a community detection algorithm discovers likely communities in a social network, while a shortest path algorithm computes the quickest way to reach a place from another, in a road network. In Domains such as social information systems, the number of edges can be in billions or trillions. Such large graphs are processed on distributed computer systems or clusters. Graph algorithms can be executed on multi-core CPUs, GPUs with thousands of cores, multi-GPU devices, and CPU+GPU clusters, depending on the size of the graph object. While programming such algorithms on heterogeneous targets, a programmer is required to deal with parallelism and and also manage explicit data communication between distributed devices. This implies that a programmer is required to learn CUDA, OpenMP, MPI, etc., and also the details of the hardware architecture. Such codes are error prone and di cult to debug. A Domain Speci c Language (DSL) which hides all the hardware details and lets the programmer concentrate only the algorithmic logic will be very useful. With this as the research goal, Falcon, graph DSL and its compiler have been developed. Falcon programs are explicitly parallel and Falcon hides all the hardware details from the programmer. Large graphs that do not t into the memory of a single device are automatically partitioned by the Falcon compiler. Another feature of Falcon is that it supports mutation of graph objects and thus enables programming dynamic graph algorithms. The Falcon compiler converts a single DSL code to heterogeneous targets such as multi-core CPUs, GPUs, multi-GPU devices, and CPU+GPU clusters. Compiled codes of Falcon match or outperform state-of-the-art graph frameworks for di erent target platforms and benchmarks.

Les systèmes temps réel (STR) sont au coeur de machines souvent jugés critiques pour lasécurité : ils en contrôlent l’exécution afin que celles-ci se comportent de manière sûre dans le contexte d’un environnement dont l’évolution peut être imprévisible. Un STR n’a d’autre alternative que de s’adapter à son environnement : sa correction dépend des temps de réponses aux stimuli de ce dernier.Il est couramment admis que le formalisme des réseaux de Petri temporels (RdPT) est adapté àla description des STR. Cependant, la modélisation de systèmes simples, ne possédant que quelquestˆaches périodiques ordonnancées de façon basique se révèle être un exercice souvent complexe.En premier lieu, la modélisation efficace d’une gamme étendue de politiques d’ordonnancementsse heurte à l’incapacité des RdPT à imposer un ordre d’apparition à des évènements concurrentssurvenant au même instant. D’autre part, les STR ont une nette tendance à être constitués de caractéristiques récurrentes, autorisant une modélisation par composants. Or les RdPT ne sont guèreadaptés à une utilisation compositionnelle un tant soit peu générale. Afin de résoudre ces deuxproblèmes, nous proposons dans cette thèse Cifre – en partenariat entre Airbus et le Laas-Cnrs– d’étendre les RdPT à l’aide de deux nouvelles relations, les relations d’inhibition et de permission,permettant de spécifier de manière plus fine les contraintes de temps.Afin de cerner un périmètre clair d’adéquation de cette nouvelle extension à la modélisation dessystèmes temps réel, nous avons défini Pola, un langage spécifique poursuivant deux objectifs :déterminer un sous-ensemble des systèmes temps réel modélisables par les réseaux de Petri temporelsà inhibitions/permissions et fournir un langage simple à la communauté temps réel dont lavérification, idéalement automatique, est assurée par construction. Sa sémantique est donnée par traduction en réseaux de Petri temporels à inhibitions/permissions. L’explorateur d’espace d’états de laboite à outils Tina a été étendu afin de permettre la vérification des descriptions Pola / Real time systems (RTS) are at the core of safety critical devices : they control thedevices’ behavior in such a way that they remain safe with regard to an unpredictable environment. ARTS has no other choices than to adapt to its environment : its correctness depends upon its responsetime to the stimuli stemming from the environment.It is widely accepted that the Time Petri nets (TPN) formalism is adapted to the description ofRTS. However, the modeling of simple systems with only a few periodic tasks scheduled according toa basic policy remains a challenge in the worst case and can be very tedious in the most favorable one.First, we put forward some limitations of TPN regarding the modeling of a wide variety of schedulingpolicies, coming from the fact that this formalism is not always capable to impose a givenorder on events whenever they happen at the same time. Moreover, RTS are usually constituted of thesame recurring features, implying a compositional modeling, but TPN are not well adapted to sucha compositional use. To solve those problems we propose in this Cifre thesis – in partnership withAirbus and the Laas-Cnrs – to extend the formalism with two new dual relations, the forbid andallow relations so that time constraints can be finely tuned.Then, to assess this new extension for modeling of real time systems, we define Pola, a specificlanguage aimed at two goals : to determine a subset of RTS which can be modeled with forbid/allowtime Petri nets and to provide a simple language to the real time community which, ideally, can bechecked automatically. Its semantics is given by translation into forbid/allow Time Petri nets. Thestate space exploration tool of the Tina toolbox have been extended so that it can model check Poladescriptions.

Ensuring code quality is crucial in software development. Not seldom, developers resort to static analysis tools to assist them in both understanding pieces of code and identifying defects or refactoring opportunities during development activities. A critical issue when defining such tools is their ability to obtain information about code. Static analysis tools depend, in general, of an intermediate program representation to identify locations that meet the conditions described in their algorithms. This perspective can be enlarged when techniques of crosscutting concerns modularization, such as aspect-oriented programming (AOP) is applied. In AOP applications, a piece of code can be systematically affected, using both static and dynamic combinations. The main goal of this dissertation is the specification and the implementation of AQL, a domain-specific language (DSL) designed to search aspect-oriented code bases. AQL is a declarative language, based on object query language (OQL), which enables the task of querying elements, relationships and program metrics to support the construction of static analysis and code searching tools for aspect oriented programs. The language was designed in two steps. First, we built a framework (AOPJungle), responsible to extract data from aspect-oriented programs. AOPJungle performs the computation of metrics, inferences and connections between the elements of the program. In the second step, we built an AQL compiler as a reference implementation. We adopted a source-to-source transformation for this step, in which an AQL query is transformed into HQL statements before being executed. In order to evaluate the reference implementation, we developed a static analysis tool for identifying refactoring opportunities in aspect-oriented programs. This tool receives a set of AQL queries to identify potential scenarios where refactoring could be applied. / Assegurar a qualidade de código é um ponto crucial durante o desenvolvimento de software. Frequentemente, os desenvolvedores recorrem às ferramentas de análise estática para auxiliá-los tanto na compreensão de código, quanto na identificação de defeitos ou de oportunidades de refatoração durante o ciclo de desenvolvimento de aplicações. Um dos pontos críticos na definição de tais ferramentas está na sua capacidade de obter informações a respeito de código. As ferramentas de análise estática dependem, em geral, de uma representação intermediária de um programa para identificar situações que atendam às condições necessárias descritas em seus algoritmos. Esse panorama se amplia com o uso de técnicas de modularização de interesses transversais, tais como a programação orientada a aspectos (POA), na qual um código pode ser afetado de forma sistêmica, por meio de combinações estáticas e dinâmicas. O principal objetivo desta dissertação é a especificação e implementação de AQL, uma DSL (linguagem específica de domínio) para a realização de busca em código orientado a aspectos. A AQL é uma linguagem declarativa, baseada em linguagem de busca em objetos (OQL) e que permite consultar elementos, relações, derivações e métricas de um programa orientado a aspectos (OA), a fim de apoiar a construção de ferramentas de análise estática e de pesquisa em código. O projeto de implementação da linguagem foi realizado em duas etapas. Primeiro, foi criado um framework (AOPJungle) para a extração de dados de programas OA. O AOPJungle além de extrair dados de programas OA, realiza a computação de métricas, inferências e ligações entre os elementos de um programa. Na segunda etapa, um compilador de referência para AQL foi construído. A abordagem adotada foi a transformação fonte a fonte, sendo uma consulta AQL transformada em uma consulta HQL (Hibernate Query Language) antes de sua execução. A fim de avaliar a implementação proposta, uma ferramenta de análise estática para identificação de oportunidades de refatoração em programas AO foi elaborada, usando a AQL para a busca de dados sobre esses programas.

De nos jours, les processus métiers permettent une automatisation croissante des tâches et des interconnexions complexes au sein du même système et entre différents systèmes, ce qui est particulièrement facilité par l'émergence des services Web. Dans ce contexte, les tâches de spécification et de vérification de la conformité pendant l’exécution deviennent particulièrement intéressantes. Dans cette thèse, on s’intéresse à deux aspects, le monitoring et la sécurité dans le contexte de l’Architecture Orienté Service (SOA). Ainsi, nous proposons une approche fondée sur le concept de vue et une plateforme qui vise le monitoring de la conformité des processus métiers pendant leur exécution. Ainsi, nous avons développé un langage de monitoring appelé BPath, qui est un langage basé sur XPath, qui offre entre autres, la possibilité de spécifier et de vérifier des propriétés de la logique temporelle linéaire et hybride, des requêtes visant à évaluer des indicateurs quantitatifs sur l’exécution d’un processus métier, ceci dans le but de détecter toute violation des règles de conformité pendant l’exécution.Une des préoccupations spécifiques du monitoring de la conformité pour les environnements basés sur SOA est la sécurité. Ainsi, nous proposons une architecture de sécurité fondée sur des langages dédiés (DSL) pour SOA. Nous avons particulièrement développé une DSL graphique pour faciliter la spécification et la génération des contrôles d’accès. Nos approches sont mises en œuvre et intégrés dans une plateforme développée dans le cadre du projet Européen COMPAS qui vise à assurer la conformité de bout en bout dans les environnements basés sur SOA. / Nowadays, business processes allow more automation of tasks and complex interconnections within the same system and across different systems, which is particularly facilitated by the emergence of Web services. In this context, the tasks of specifying and checking compliance at runtime become particularly challenging.In this thesis, our goal is twofold: monitoring and security in the context of Service Oriented Architecture (SOA). Thus, we proposed a view-based monitoring approach and a framework that target monitoring of business process compliance at runtime. Our monitoring framework aims to offer an easy way to specify properties to be monitored and to facilitate its integration with SOA based environments. Thus, we have developed a new monitoring language called BPath, which is an XPath-based language that offers among others, the ability to express and to check temporal and hybrid logic properties at runtime, making the execution of business processes visible by expressing and evaluating quantitative indicators, in order to detect any compliance violation at runtime. A specific compliance monitoring concern in SOA based environment is security, which is also an important aspect for companies willing to give access to some of their resources over the Web. Thus, we proposed a domain specific language (DSL) based architecture for ensuring security in SOA environments. We particularly focused on access control by proposing a graphical language to facilitate the specification and generation of access control policies.Our approaches are implemented and integrated within a complete end to end compliance framework developed within the COMPAS project.