Search

International audience; Model driven engineering advocates the separation of concerns during the design time of a system, which leads to the creation of several different models, using several different syntaxes. However, to reason on the overall system, we need to compose these models. Unfortunately, composition of models is done in an ad hoc way, preventing comparison, capitalisation and reuse of the composition operators. In order to improve comprehension and allow comparison of merging and weaving operators, we use category theory to propose a unified framework to formally define merging and weaving of models. We successfully use this framework to compare them, both through the way they are transformed in the formalism, and through several properties, such as completeness or non-redundancy. Finally, we validate this framework by checking that it correctly identifies three tools as performing merging or weaving of models.

We propose, in this master internship report, a unified formal framework for model composition. Indeed, there are various operators of model composition de fined each in its own framework. Defining a unified framework for model composition allows to compare them, to reuse them and to capitalize knowledge. Defining a formal framework allows to consider the system that these compositions represent in its entirety. This framework is built through category theory and the notion of pushout. More precisely, a category of models is defined, and the pushout construction in this category described. Then, this pushout is used to formalize several model composition operators, including model merge, model transformation and aspect weaving.

National audience; LEOPAR est un analyseur syntaxique fondé sur le formalisme des grammaires d'interaction, intégré dans une chaîne d'outils visant avant tout à la précision et la couverture linguistique. À l'aide de XMG, il est possible de construire des grammaires de façon semi-automatique à partir de connaissances linguistiques. Ces grammaires sont ensuite ancrées à l'aide de lexiques indépendants d'un quelconque formalisme. Les grammaires ancrées sont enfin utilisées par LEOPAR pour analyser des corpus bruts et produire des sorties sous forme d'arbres syntagmatiques ou de structures de dépendances.

National audience; This article proposes a method to extract dependency structures from phrase-structure level parsing with Interaction Grammars. Interaction Grammars are a formalism which expresses interactions among words using a polarity system. Syntactical composition is led by the saturation of polarities. Interactions take place between constituents, but as grammars are lexicalized, these interactions can be translated at the level of words. Dependency relations are extracted from the parsing process: every dependency is the consequence of a polarity saturation. The dependency relations we obtain can be seen as a refinement of the usual dependency tree. Generally speaking, this work sheds new light on links between phrase structure and dependency parsing.

This article proposes a method to extract dependency structures from phrase-structure level parsing with Interaction Grammars. Interaction Grammars are a formalism which expresses interactions among words using a polarity system. Syntactical composition is led by the saturation of polarities. Interactions take place between constituents, but as grammars are lexicalized, these interactions can be translated at the level of words. Dependency relations are extracted from the parsing process: every dependency is the consequence of a polarity saturation. The dependency relations we obtain can be seen as a refinement of the usual dependency tree. Generally speaking, this work sheds new light on links between phrase structure and dependency parsing.