Title: Toward concept visualization through image generation Author: Tien Dat Nguyen Department: Institute of Formal and Applied Linguistics Supervisors: Pavel Pecina (Charles University in Prague), Angeliki Lazaridou, Raffaella Bernardi, Marco Baroni (University of Trento), Abstract: Computational linguistic and computer vision have a common way to embed the semantics of linguistic/visual units through vector representation. In addition, high-quality semantic representations can be effectively constructed thanks to recent advances in neural network methods. Nevertheless, the under- standing of these representations remains limited, so they need to be assessed in an intuitive way. Cross-modal mapping is mapping between vector semantic embedding of words and the visual representations of the corresponding objects from images. Inverting image representation involves learning an image inversion of visual vectors (SIFT, HOG and CNN features) to reconstruct the original one. The goal of this project is to build a complete pipeline, in which word represen- tations are transformed into image vectors using cross modal mapping and these vectors are projected to pixel space using inversion. This suggests that there might be a groundbreaking way to inspect and evaluate the semantics encoded in word representations by...

As transformações de modelos são um ponto vital para os enfoques de desenvolvimento dirigido por modelos, incluindo a MDA. No presente trabalho apresenta-se alguns dos conceitos da MDA relacionados a transformações de modelos. São vistos os conceitos de PIM, PSM, transformações, mapeamentos e marcas. Estes conceitos são analisados no contexto de três ferramentas MDA, a OptimalJ, a AndroMDA e a ArcStyler, para verificar se as ferramentas os implementam de acordo com a MDA. Também é analisado como e quais transformações e mapeamentos de modelos estão sendo implementados pelas ferramentas, através da comparação qualitativa de definições de conceitos e de exemplos do uso destes. Além de verificar os conceitos relacionados às transformações, faz-se um estudo de caso qualitativo para analisar as transformações e mapeamentos providos pelas ferramentas e como as ferramentas trabalham com eles. Mostram-se as dificuldades de se criar um modelo PIM do sistema do estudo de caso e adaptá-lo para cada uma das ferramentas. / Model transformations are a vital point for model driven development, including MDA. In the present essay we introduce some of the MDA concepts related with model transformations. We present the concept of PIM, PSM, transformations, mappings and marks. These concepts are analyzed in the context of three MDA tools, OptimalJ, AndroMDA and ArcStyler, to verify if the tools implement the concepts according to MDA. We also analyze how and which model transformations and model mappings are being implemented by the tools, through a qualitative comparison of the definition of these concepts and using some examples of their intended use. In addition to verifying the concepts related to transformations, we create a qualitative case study to analyze the model transformations and model mappings provided by the tools, and how the tools work with them. We show a few of the problems faced when creating a PIM model of the case study\'s system and the adaptations it required for each tool.

As transformações de modelos são um ponto vital para os enfoques de desenvolvimento dirigido por modelos, incluindo a MDA. No presente trabalho apresenta-se alguns dos conceitos da MDA relacionados a transformações de modelos. São vistos os conceitos de PIM, PSM, transformações, mapeamentos e marcas. Estes conceitos são analisados no contexto de três ferramentas MDA, a OptimalJ, a AndroMDA e a ArcStyler, para verificar se as ferramentas os implementam de acordo com a MDA. Também é analisado como e quais transformações e mapeamentos de modelos estão sendo implementados pelas ferramentas, através da comparação qualitativa de definições de conceitos e de exemplos do uso destes. Além de verificar os conceitos relacionados às transformações, faz-se um estudo de caso qualitativo para analisar as transformações e mapeamentos providos pelas ferramentas e como as ferramentas trabalham com eles. Mostram-se as dificuldades de se criar um modelo PIM do sistema do estudo de caso e adaptá-lo para cada uma das ferramentas. / Model transformations are a vital point for model driven development, including MDA. In the present essay we introduce some of the MDA concepts related with model transformations. We present the concept of PIM, PSM, transformations, mappings and marks. These concepts are analyzed in the context of three MDA tools, OptimalJ, AndroMDA and ArcStyler, to verify if the tools implement the concepts according to MDA. We also analyze how and which model transformations and model mappings are being implemented by the tools, through a qualitative comparison of the definition of these concepts and using some examples of their intended use. In addition to verifying the concepts related to transformations, we create a qualitative case study to analyze the model transformations and model mappings provided by the tools, and how the tools work with them. We show a few of the problems faced when creating a PIM model of the case study\'s system and the adaptations it required for each tool.

Uma das principais atividades dos enfoques de desenvolvimento de software centrados em modelos, como por exemplo a Arquitetura Dirigida por Modelo (Model Driven Architecture - MDA), é o processo de transformação de modelos. Geralmente, um passo preliminar para a transformação dos modelos é o mapeamento dos elementos do meta-modelo fonte nos elementos do meta-modelo alvo. Este trabalho apresenta uma aplicação de uma técnica de mapeamento de modelos denominada tecelagem de modelos. Esta técnica permite ao usuário definir a semântica das ligações estabelecidas entre os elementos do meta-modelo fonte e os elementos do meta-modelo alvo. A semântica é definida através de tipos fortes associados às ligações. O presente trabalho analisa, através de dois experimentos, alguns aspectos da geração de modelos de transformação de modelos no arcabouço MDA, utilizando a técnica de tecelagem de modelos. A análise utiliza duas alternativas de especificação de transformação de modelos a título de comparação: a que usa somente uma linguagem de especificação de modelos de transformação e a que usa a técnica de tecelagem de modelos. Os aspectos analisados são: a reutilização de trechos de código escritos na linguagem de geração de especificações de transformação e a reutilização de decisões de projeto (design) no mapeamento entre dois meta-modelos distintos. / One of the main activities of the model-centric approaches of software development, as for example the Model Driven Architecture (MDA), is the process of model transformation. Usually, a preliminary step for model transformation is the mapping of source metamodel elements into target metamodel elements. This work presents an application of a technique for model mapping called model weaving. This technique allows the user to define the semantics of the links binding source metamodel elements and target metamodel elements. The semantics is defined through types associated to links. The work analyzes, through two experiments, some aspects of the generation of models transformation in the MDA framework, using a technique known as model weaving. The analysis, for comparison purposes, uses two techniques of model transformation specification: one using only a specification language for model transformation specification and another using model weaving. The analyzed aspects are: the reuse of pieces of code written in the language that generates the transformation of specifications and the reuse of design decisions in the mapping between two distinct metamodels.

Uma das principais atividades dos enfoques de desenvolvimento de software centrados em modelos, como por exemplo a Arquitetura Dirigida por Modelo (Model Driven Architecture - MDA), é o processo de transformação de modelos. Geralmente, um passo preliminar para a transformação dos modelos é o mapeamento dos elementos do meta-modelo fonte nos elementos do meta-modelo alvo. Este trabalho apresenta uma aplicação de uma técnica de mapeamento de modelos denominada tecelagem de modelos. Esta técnica permite ao usuário definir a semântica das ligações estabelecidas entre os elementos do meta-modelo fonte e os elementos do meta-modelo alvo. A semântica é definida através de tipos fortes associados às ligações. O presente trabalho analisa, através de dois experimentos, alguns aspectos da geração de modelos de transformação de modelos no arcabouço MDA, utilizando a técnica de tecelagem de modelos. A análise utiliza duas alternativas de especificação de transformação de modelos a título de comparação: a que usa somente uma linguagem de especificação de modelos de transformação e a que usa a técnica de tecelagem de modelos. Os aspectos analisados são: a reutilização de trechos de código escritos na linguagem de geração de especificações de transformação e a reutilização de decisões de projeto (design) no mapeamento entre dois meta-modelos distintos. / One of the main activities of the model-centric approaches of software development, as for example the Model Driven Architecture (MDA), is the process of model transformation. Usually, a preliminary step for model transformation is the mapping of source metamodel elements into target metamodel elements. This work presents an application of a technique for model mapping called model weaving. This technique allows the user to define the semantics of the links binding source metamodel elements and target metamodel elements. The semantics is defined through types associated to links. The work analyzes, through two experiments, some aspects of the generation of models transformation in the MDA framework, using a technique known as model weaving. The analysis, for comparison purposes, uses two techniques of model transformation specification: one using only a specification language for model transformation specification and another using model weaving. The analyzed aspects are: the reuse of pieces of code written in the language that generates the transformation of specifications and the reuse of design decisions in the mapping between two distinct metamodels.

No description available.

XML is an emerging standard format for data on the Web as well as in business applications. In order to store and access this information in an efficient manner, database technology must be utilized. A relational database system, the most established and mature technology for query processing and storage, creates a strong foundation for such an XML data management system. However, while relational databases are based on SQL queries, the original user queries are written in XQuery, an XML query language. This XML query language has support for order-sensitive queries as XML is an order-sensitive markup language. A major problem has been discovered with loading XML in a relational database. That problem is the lack of native SQL support for and management of order handling. While XQuery has order and positional support, SQL does not have the same support. For example, individuals who were viewing XML information about music albums would have a hard time querying for the first three songs of a track list from a relational backend. Mapping XML documents to relational backends also proves hard as the data models (hierarchical elements versus flat tables) are so different. For these reasons, and other purposes, the Rainbow System is being developed at WPI as a system that bridges XML data and relational data. This thesis in particular deals with the algebra operators that affect order, order sensitive loading and mapping of XML documents, and the pushdown of order handling into SQL-capable query engines. The contributions of the thesis are the order-sensitive rewrite rules, new XML to relational mappings with different order styles, order-sensitive template-driven SQL generation, and a proposed metadata table for order-sensitive information. A system that implements these proposed techniques with XQuery as the XML query language and Oracle as the backend relational storage system has been developed. Experiments were created to measure execution time based on various factors. First, scalability of the system as backend data set size grows is studied. Second, scalability of the system as results returned from the database grows, and finally, query execution times with different loading types are explored. The experimental results are encouraging. Query execution with the relational backend proves to be much faster than native execution within the Rainbow system. These results confirm the practical utility of our proposed order-sensitive XQuery execution solution over relational data.

Mit der stetigen Konsolidierung der Web Service Standards vollzieht sich in der Diskussion zu serviceorientierten Architekturen ein Wechsel weg von der Implementierung und dem Deployment von Services hin zum Service Management. Die Indikatoren: Anzahl der Standardisierungsanfragen und die Menge großer Forschungsprogramme in diesem Bereich sind evident für einen wachsenden Bedarf an Managementmethoden für die Abbildung von Geschäftsanforderungen auf Servicekompositionen. Die Verwendung von semiformalen Modellen sowie die verteilte Entwicklung von SOA Services machen eine an Konventionen orientierte konstruktive Einschränkung der Freiheitsgrade bei der Prozess- und Servicemodellierung notwendig. In dieser Arbeit wird eine generische Methode vorgestellt, welche durch die Einführung von „Beschreibungsrahmen“ (Description Kits) sowohl eine Einschränkung der Modellierungsfreiheit in Bezug auf natürlichsprachliche Aspekte in fachkonzeptuellen Modellen erlaubt, als auch eine restriktive Benutzung bestehender Modellierungssprachen ermöglicht. Als Anwendungsfall wird die Konfiguration serviceorientierter Architekturen diskutiert. - (Die Dissertation ist veröffentlicht im Logos Verlag Berlin GmbH, http://www.logos-verlag.de, ISBN: 978-3-8325-2402-9) / With Service-oriented Architectures (SOA) companies are facing a paradigm shift towards services as the key unit of their enterprise architectures. An exploitation of the potential of Service-oriented Architectures is up to the applicability of semi-formal business models for a transformation into service compositions. However, several conceptual shortcomings prevent semi-formal models from being a suitable resource of an automated transformation process. Problems result from a multitude of modeling aims, objects, and procedures. Enhancements in order to solve these problems require both systematic preparation, and a methodical approach. This thesis provides a modeling method for business process-oriented SOA design using a meta-model based approach. An integrated modeling environment gives an overview of business requirements and allows appropriate assignment of SOA services to the business functions at the same time. The research refers to the example of process-oriented analysis and optimization of administrative procedures in German universities.

Mit der stetigen Konsolidierung der Web Service Standards vollzieht sich in der Diskussion zu serviceorientierten Architekturen ein Wechsel weg von der Implementierung und dem Deployment von Services hin zum Service Management. Die Indikatoren: Anzahl der Standardisierungsanfragen und die Menge großer Forschungsprogramme in diesem Bereich sind evident für einen wachsenden Bedarf an Managementmethoden für die Abbildung von Geschäftsanforderungen auf Servicekompositionen. Die Verwendung von semiformalen Modellen sowie die verteilte Entwicklung von SOA Services machen eine an Konventionen orientierte konstruktive Einschränkung der Freiheitsgrade bei der Prozess- und Servicemodellierung notwendig. In dieser Arbeit wird eine generische Methode vorgestellt, welche durch die Einführung von „Beschreibungsrahmen“ (Description Kits) sowohl eine Einschränkung der Modellierungsfreiheit in Bezug auf natürlichsprachliche Aspekte in fachkonzeptuellen Modellen erlaubt, als auch eine restriktive Benutzung bestehender Modellierungssprachen ermöglicht. Als Anwendungsfall wird die Konfiguration serviceorientierter Architekturen diskutiert. - (Die Dissertation ist veröffentlicht im Logos Verlag Berlin GmbH, http://www.logos-verlag.de, ISBN: 978-3-8325-2402-9) / With Service-oriented Architectures (SOA) companies are facing a paradigm shift towards services as the key unit of their enterprise architectures. An exploitation of the potential of Service-oriented Architectures is up to the applicability of semi-formal business models for a transformation into service compositions. However, several conceptual shortcomings prevent semi-formal models from being a suitable resource of an automated transformation process. Problems result from a multitude of modeling aims, objects, and procedures. Enhancements in order to solve these problems require both systematic preparation, and a methodical approach. This thesis provides a modeling method for business process-oriented SOA design using a meta-model based approach. An integrated modeling environment gives an overview of business requirements and allows appropriate assignment of SOA services to the business functions at the same time. The research refers to the example of process-oriented analysis and optimization of administrative procedures in German universities.