abstract: Knowledge representation and reasoning is a prominent subject of study within the field of artificial intelligence that is concerned with the symbolic representation of knowledge in such a way to facilitate automated reasoning about this knowledge. Often in real-world domains, it is necessary to perform defeasible reasoning when representing default behaviors of systems. Answer Set Programming is a widely-used knowledge representation framework that is well-suited for such reasoning tasks and has been successfully applied to practical domains due to efficient computation through grounding--a process that replaces variables with variable-free terms--and propositional solvers similar to SAT solvers. However, some domains provide a challenge for grounding-based methods such as domains requiring reasoning about continuous time or resources. To address these domains, there have been several proposals to achieve efficiency through loose integrations with efficient declarative solvers such as constraint solvers or satisfiability modulo theories solvers. While these approaches successfully avoid substantial grounding, due to the loose integration, they are not suitable for performing defeasible reasoning on functions. As a result, this expressive reasoning on functions must either be performed using predicates to simulate the functions or in a way that is not elaboration tolerant. Neither compromise is reasonable; the former suffers from the grounding bottleneck when domains are large as is often the case in real-world domains while the latter necessitates encodings to be non-trivially modified for elaborations. This dissertation presents a novel framework called Answer Set Programming Modulo Theories (ASPMT) that is a tight integration of the stable model semantics and satisfiability modulo theories. This framework both supports defeasible reasoning about functions and alleviates the grounding bottleneck. Combining the strengths of Answer Set Programming and satisfiability modulo theories enables efficient continuous reasoning while still supporting rich reasoning features such as reasoning about defaults and reasoning in domains with incomplete knowledge. This framework is realized in two prototype implementations called MVSM and ASPMT2SMT, and the latter was recently incorporated into a non-monotonic spatial reasoning system. To define the semantics of this framework, we extend the first-order stable model semantics by Ferraris, Lee and Lifschitz to allow "intensional functions" and provide analyses of the theoretical properties of this new formalism and on the relationships between this and existing approaches. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2016

abstract: Several physical systems exist in the real world that involve continuous as well as discrete changes. These range from natural dynamic systems like the system of a bouncing ball to robotic dynamic systems such as planning the motion of a robot across obstacles. The key aspects of effectively describing such dynamic systems is to be able to plan and verify the evolution of the continuous components of the system while simultaneously maintaining critical constraints. Developing a framework that can effectively represent and find solutions to such physical systems prove to be highly advantageous. Both hybrid automata and action languages are formal models for describing the evolution of dynamic systems. The action language C+ is a rich and expressive language framework to formalize physical systems, but can be used only with physical systems in the discrete domain and is limited in its support of continuous domain components of such systems. Hybrid Automata is a well established formalism used to represent such complex physical systems at a theoretical level, however it is not expressive enough to capture the complex relations between the components of the system the way C+ does. This thesis will focus on establishing a formal relationship between these two formalisms by showing how to succinctly represent Hybrid Automata in an action language which in turn is defined as a high-level notation for answer set programming modulo theories (ASPMT) --- an extension of answer set programs in the first-order level. Furthermore, this encoding framework is shown to be more effective and expressive than Hybrid Automata by highlighting its ability in allowing states of a hybrid transition system to be defined by complex relations among components that would otherwise be abstracted away in Hybrid Automata. The framework is further realized in the implementation of the system CPLUS2ASPMT, which takes advantage of state of the art ODE(Ordinary Differential Equations) based SMT solver dReal to provide support for ODE based evolution of continuous components of a dynamic system. / Dissertation/Thesis / Masters Thesis Computer Science 2017

abstract: Image Understanding is a long-established discipline in computer vision, which encompasses a body of advanced image processing techniques, that are used to locate (“where”), characterize and recognize (“what”) objects, regions, and their attributes in the image. However, the notion of “understanding” (and the goal of artificial intelligent machines) goes beyond factual recall of the recognized components and includes reasoning and thinking beyond what can be seen (or perceived). Understanding is often evaluated by asking questions of increasing difficulty. Thus, the expected functionalities of an intelligent Image Understanding system can be expressed in terms of the functionalities that are required to answer questions about an image. Answering questions about images require primarily three components: Image Understanding, question (natural language) understanding, and reasoning based on knowledge. Any question, asking beyond what can be directly seen, requires modeling of commonsense (or background/ontological/factual) knowledge and reasoning. Knowledge and reasoning have seen scarce use in image understanding applications. In this thesis, we demonstrate the utilities of incorporating background knowledge and using explicit reasoning in image understanding applications. We first present a comprehensive survey of the previous work that utilized background knowledge and reasoning in understanding images. This survey outlines the limited use of commonsense knowledge in high-level applications. We then present a set of vision and reasoning-based methods to solve several applications and show that these approaches benefit in terms of accuracy and interpretability from the explicit use of knowledge and reasoning. We propose novel knowledge representations of image, knowledge acquisition methods, and a new implementation of an efficient probabilistic logical reasoning engine that can utilize publicly available commonsense knowledge to solve applications such as visual question answering, image puzzles. Additionally, we identify the need for new datasets that explicitly require external commonsense knowledge to solve. We propose the new task of Image Riddles, which requires a combination of vision, and reasoning based on ontological knowledge; and we collect a sufficiently large dataset to serve as an ideal testbed for vision and reasoning research. Lastly, we propose end-to-end deep architectures that can combine vision, knowledge and reasoning modules together and achieve large performance boosts over state-of-the-art methods. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2018

Seres humanos são capazes de desenvolver complexas estruturas de conhecimento que podem ser utilizadas de modo flexível para lidar com o ambiente de maneira apropriada. Estas estruturas de conhecimento constituem um núcleo que suporta processos cognitivos, tais como a percepção, a categorização, o planejamento, etc. A Inteligência Artificial, enquanto área de investigação, ocupa-se de desenvolver meios que viabilizem a reprodução destas capacidades cognitivas em agentes artificiais. Por este motivo, a investigação de abordagens que permitam a representação de conhecimento de um modo flexível se revela altamente relevante. Com o objetivo de superar algumas das limitações típicas da teoria clássica, que é adotada por várias abordagens propostas na Inteligência Artificial, este trabalho propõe um arcabouço cognitivamente inspirado para representação de conhecimento e raciocínio que integra aspectos de três diferentes teorias cognitivas a respeito de como conceitos são representados na cognição humana: teoria clássica, teoria do protótipo e teoria do exemplar. O arcabouço resultante é capaz de suportar a composicionalidade, a tipicalidade, a representação de instâncias atípicas dos conceitos, e a representação da variabilidade de indivíduos classificados por cada conceito. Consequentemente, o arcabouço proposto também suporta raciocínio lógico e baseado em similaridade. As principais contribuições deste trabalho são a concepção teórica e a formalização de um arcabouço cognitivamente inspirado para representação de conhecimento e raciocínio. Uma outra contribuição deste trabalho é uma abordagem de raciocínio para classificação que utiliza a abordagem de representação de conhecimento proposta. Além disso, este trabalho também apresenta duas abordagens para seleção de exemplares representativos de cada conceito e uma abordagem para extração de protótipos de conceitos. Nesta tese também é apresentado um sistema para interpretação automática de processos deposicionais que adota o arcabouço proposto. Experimentos realizados em uma tarefa de classificação sugerem que o arcabouço proposto é capaz de oferecer classificações mais informativas que as oferecidas por uma abordagem puramente clássica. / Human beings can develop complex knowledge structures that can be used for dealing with the environment in suitable ways. These knowledge structures constitute a core that supports several cognitive processes, such as perception, categorization, planning, etc. The Artificial Intelligence, as a research field, aims at developing approaches for mimicking these cognitive capabilities in machines. Due to this, it is important to investigate approaches that allow representing the knowledge in flexible ways. In order to overcome some limitations of the classical theory of knowledge representation, which is adopted by several approaches proposed in the Artificial Intelligence field, this work proposes a cognitively-inspired framework for knowledge representation and reasoning which integrates aspects from three different cognitive theories about concept representation in the human cognition: classical theory, prototype theory and exemplar theory. The resulting framework can support compositionality, typicality, representation of atypical instances of concepts, and representation of the variability of the individuals classified by each concept. Consequently, the proposed framework also supports logical reasoning and similarity-based reasoning. The main contributions of this work are the formalization of a cognitively-inspired framework for knowledge representation and reasoning, two approaches for selecting representative exemplars of each concept and an approach of reasoning for classification that integrates logical reasoning and similarity-based reasoning and that is supported by definitions, prototypes and exemplars of concepts. This thesis also presents a system for automatic interpretation of depositional processes application that adopts the proposed framework. The experiments, which were performed on a classification task, suggest that the proposed framework provides classifications that are more informative than the ones provided by a classical approach.

O projeto de produtos e sistemas técnicos complexos requer a compreensão em nível de sistemas e subsistemas para formular soluções eficientes e integradas ao seu contexto. Para auxiliar esta compreensão, o conhecimento de projeto deve ser representado utilizando níveis adequados de abstração de acordo com a fase do projeto. A fase de projeto conceitual requer tipos de representação capazes de atingir um alto nível de abstração para a exploração de conceitos que conduzam a soluções criativas. O paradigma da orientação a objetos, que é fundamentado pela abstração, faz parte da engenharia de software, mas também pode ser aplicado para o projeto de artefatos físicos porque permite a representação dos elementos do mundo real através de uma linguagem simples, acessível e com alto nível de abstração. Ademais, o paradigma da orientação a objetos permite a reutilização do conhecimento de projeto devido à sua capacidade de estruturar a informação em um formato adequado para isto. O presente trabalho propõe um framework para representar o conhecimento de projeto de produto aplicando o paradigma da orientação a objeto. Inicialmente, foram identificados os elementos conceituais da tese e suas relações, para na sequência definir o modelo do framework e o seu método de aplicação O framework utiliza uma linguagem de representação diagramática que pode evoluir desde um mapa mental, com elementos diversificados e pouco ordenados, até uma rede estruturada de classes e relacionamentos em um modelo de classes. Um modelo de classes pode concentrar conhecimento sobre o projeto, servindo como uma estrutura geral que conecta e relaciona diferentes blocos de informação associados aos produtos e sistemas que estão sendo elaborados. A verificação da aplicabilidade do framework foi realizada por especialistas da área de design mediante o desenvolvimento de um projeto de produto em nível conceitual e do preenchimento de questionário de avaliação. / The design of complex technical products requires understanding at the system and subsystem level to formulate efficient and integrated solutions to their context. To support this understanding, the project knowledge can be represented using appropriate levels of abstraction according to the project phase. The conceptual design phase requires types of representation that reach a high level of abstraction for the exploration of concepts that lead to creative solutions. The object-oriented paradigm is based on abstraction and is part of software engineering, but can also be applied to the design of physical artifacts because it allows the representation of realworld elements through simple, accessible and in high-level abstraction language. In addition, the object orientation paradigm supports the reusability of project knowledge due to its capacity to structure the information in patterns. The present work proposes a framework to represent product design knowledge using the objectoriented paradigm First, the conceptual elements of the thesis and their relationships were identified, after; the framework model and their method of application were constructed. The framework uses a diagrammatic representation language in which a mental map, with diversified and unordered elements, can progress into a structured network of classes and relationships in a class model. A class model can focus knowledge about the project, serving as a general structure that connects and relates different blocks of information associated with the products and systems being developed. The verification of the applicability of the framework was carried out by specialists in the design area through the development of a product design at conceptual level and the answering an evaluation questionnaire.

abstract: Biological organisms are made up of cells containing numerous interconnected biochemical processes. Diseases occur when normal functionality of these processes is disrupted, manifesting as disease symptoms. Thus, understanding these biochemical processes and their interrelationships is a primary task in biomedical research and a prerequisite for activities including diagnosing diseases and drug development. Scientists studying these interconnected processes have identified various pathways involved in drug metabolism, diseases, and signal transduction, etc. High-throughput technologies, new algorithms and speed improvements over the last decade have resulted in deeper knowledge about biological systems, leading to more refined pathways. Such pathways tend to be large and complex, making it difficult for an individual to remember all aspects. Thus, computer models are needed to represent and analyze them. The refinement activity itself requires reasoning with a pathway model by posing queries against it and comparing the results against the real biological system. Many existing models focus on structural and/or factoid questions, relying on surface-level information. These are generally not the kind of questions that a biologist may ask someone to test their understanding of biological processes. Examples of questions requiring understanding of biological processes are available in introductory college level biology text books. Such questions serve as a model for the question answering system developed in this thesis. Thus, the main goal of this thesis is to develop a system that allows the encoding of knowledge about biological pathways to answer questions demonstrating understanding of the pathways. To that end, a language is developed to specify a pathway and pose questions against it. Some existing tools are modified and used to accomplish this goal. The utility of the framework developed in this thesis is illustrated with applications in the biological domain. Finally, the question answering system is used in real world applications by extracting pathway knowledge from text and answering questions related to drug development. / Dissertation/Thesis / Ph.D. Computer Science 2014

abstract: Modeling dynamic systems is an interesting problem in Knowledge Representation (KR) due to their usefulness in reasoning about real-world environments. In order to effectively do this, a number of different formalisms have been considered ranging from low-level languages, such as Answer Set Programming (ASP), to high-level action languages, such as C+ and BC. These languages show a lot of promise over many traditional approaches as they allow a developer to automate many tasks which require reasoning within dynamic environments in a succinct and elaboration tolerant manner. However, despite their strengths, they are still insufficient for modeling many systems, especially those of non-trivial scale or that require the ability to cope with exceptions which occur during execution, such as unexpected events or unintended consequences to actions which have been performed. In order to address these challenges, a theoretical framework is created which focuses on improving the feasibility of applying KR techniques to such problems. The framework is centered on the action language BC+, which integrates many of the strengths of existing KR formalisms, and provides the ability to perform efficient reasoning in an incremental fashion while handling exceptions which occur during execution. The result is a developer friendly formalism suitable for performing reasoning in an online environment. Finally, the newly enhanced Cplus2ASP 2 is introduced, which provides a number of improvements over the original version. These improvements include implementing BC+ among several additional languages, providing enhanced developer support, and exhibiting a significant performance increase over its predecessors and similar systems. / Dissertation/Thesis / M.S. Computer Science 2014

Desde o final da década de 1990 existe um interesse crescente na aplicação de técnicas de planejamento automático em IA para resolver problemas reais de engenharia. Além das características dos problemas acadêmicos, tais como a necessidade de raciocinar sobre as ações, problemas reais requerem elicitação, engenharia e gerenciamento detalhado do conhecimento do domínio. Para tais aplicações reais, um processo de design sistemático é necessário onde as ferramentas de Engenharia do Conhecimento e de Requisitons têm um papel fundamental. Esforços acadêmicos recentes na área da Engenharia do Conhecimento em planejamento automático vêm desenvolvido ferramentas e técnicas de apoio ao processo de design de modelos do conhecimento. Porém, dada a natural incompletude do conhecimento, experiência prática em aplicações reais, como por exemplo exploração do espaço, tem mostrado que, mesmo com um processo disciplinado de design, requisitos de pontos de vista diferente (por exemplo, especialistas, usuários e patrocinadores) ainda surgem após a análise, geração e execução de planos. A tese central deste texto é que uma fase de análise de pós-design para o desenvolvimento de aplicações de planejamento em IA resulta em modelos do conhecimento mais ricos e, conseqüentemente, aumenta a qualidade dos planos gerados e a performance dos planejadores automáticos. Neste texto, nós investigamos como os conhecimentos e requisitos ocultos podem ser adquiridos e reutilizados durante a fase de análise de plans (posterior ao design do modelo) e como estes conhecimentos afetam o desempenho do processo de planejamento automático. O texto descreve um framework de post-design chamado postDAM que combina (1) uma ferramenta de engenharia de conhecimento para a aquisição de requisitos e avaliação do plano, (2) um ambiente de prototipagem virtual para a análise e simulação de planos, (3) um sistema de banco de dados para armazenamento de avaliações de planos, e (4) um sistema de raciocínio ontológico para o re-uso e descoberta de conhecimento sobre o domínio. Com o framework postDAM demonstramos que a análise de pós-design auxilia a descoberta de requisitos ocultos e orienta o ciclo de refinamento do modelo. Este trabalho apresenta três estudos de caso com domínios conhecidos na literatura e oito planejadores do estado da arte. Nossos resultados demonstram que melhorias significativas na qualidade do plano e um aumento na velocidade dos planejadores de até três ordens de grandeza pode ser alcançada através de um processo disciplinado e cuidados de pós-design. Nós demonstramos também que rationales provenientes dos usuários capturados durante as avaliações de planos podem ser úteis e reutilizáveis em novas avaliações de plano e em novos projetos. Nós argumentamos que esse processo de pós-design é fundamental para a implantação da tecnologia de planejamento automático em aplicações do mundo real. Até onde sabemos, este é o primeiro trabalho que investiga a análise de pós-design em aplicações de planejamento automático da IA. / Since the end of the 1990s there has been an increasing interest in the application of AI planning techniques to solve real-life problems. In addition to characteristics of academic problems, such as the need to reason about actions, real-life problems require detailed knowledge elicitation, engineering, and management. A systematic design process in which Knowledge and Requirements Engineering techniques and tools play a fundamental role is necessary in such applications. Research on Knowledge Engineering for planning and scheduling has created tools and techniques to support the design process of planning domain models. However, given the natural incompleteness of the knowledge, practical experience in real applications such as space exploration has shown that, even with a disciplined process of design, requirements from different viewpoints (e.g. stakeholders, experts, users) still emerge after plan generation, analysis and execution. The central thesis of this dissertation is that an post-design analysis phase in the development of AI planning applications leads to richer knowledge models and, consequently, to high-performance and high-quality plans. In this dissertation, we investigate how hidden knowledge and requirements can be acquired and re-used during a plan analysis phase that follows model design and how they affect planning performance. We describe a post-design framework called postDAM that combines (1) a knowledge engineering tool for requirements acquisition and plan evaluation, (2) a virtual prototyping environment for the analysis and simulation of plans, (3) a database system for storing plan evaluations, and (4) an ontological reasoning system for knowledge re-use and discovery. Our framework demonstrates that post-design analysis supports the discovery of missing requirements and guides the model refinement cycle. We present three case studies using benchmark domains and eight state-of-the-art planners. Our results demonstrate that significant improvements in plan quality and an increase in planning speed of up to three orders of magnitude can be achieved through a careful post-design process. We also demonstrate that rationales captured during plan evaluations from users can be useful and reusable in further plan evaluations and in new application designs. We argue that such a post-design process is critical for deployment of planning technology in real-world applications. To our knowledge, this is the first work that investigate post-design analysis for AI planning applications.

As Redes Bayesianas constituem um modelo computacional adequado para a realização de inferências probabilísticas em domínios que envolvem a incerteza. O raciocínio diagnóstico médico pode ser caracterizado como um ato de inferência probabilística em um domínio incerto, onde a elaboração de hipóteses diagnósticas é representada pela estratificação de doenças em função das probabilidades a elas associadas. A presente dissertação faz uma pesquisa sobre a metodologia para construção/validação de redes bayesianas voltadas à área médica, e utiliza estes conhecimentos para o desenvolvimento de uma rede probabilística para o auxílio diagnóstico da Insuficiência Cardíaca (IC). Esta rede bayesiana, implementada como parte do sistema SEAMED/AMPLIA, teria o papel de alerta para o diagnóstico e tratamento precoce da IC, o que proporcionaria uma maior agilidade e eficiência no atendimento de pacientes portadores desta patologia. / Bayesian networks (BN) constitute an adequate computational model to make probabilistic inference in domains that involve uncertainty. Medical diagnostic reasoning may be characterized as an act of probabilistic inference in an uncertain domain, where diagnostic hypotheses elaboration is represented by the stratification of diseases according to the related probabilities. The present dissertation researches the methodology used in the construction/validation of Bayesian Networks related to the medical field, and makes use of this knowledge for the development of a probabilistic network to aid in the diagnosis of Heart Failure (HF). This BN, implemented as part of the SEAMED/AMPLIA System, would engage in the role of alerting for early diagnosis and treatment of HF, which could provide faster and more efficient healthcare of patients carrying this pathology.

Em domínios com forte conteúdo visual, a interpretação de imagens por raciocínio visual pode ser mais eficaz na solução de problemas do que a interpretação de dados puramente textuais ou numéricos. No entanto, a representação do conhecimento visual é difícil de ser realizada por tratar-se de um conhecimento implícito para o observador. As ontologias de representação possibilitam a criação de estruturas para auxiliar na captura desse tipo de conhecimento, de forma a atribuir uma representação simbólica e significado semântico ao que está sendo visualizado. A formalização do conhecimento visual permite a sua utilização em processos de inferência, resultando na interpretação automática da imagem. O objetivo deste trabalho é a definição de construtos ontológicos que permitam descrever aspectos visuais presentes em uma imagem, com ênfase na atenção visual mais do que nos aspectos físicos dos objetos. Esses aspectos visuais são associados aos objetos físicos da imagem bem como aos objetos descritos no nível do conhecimento de domínio. Para cada um dos níveis foi definida uma ontologia de representação, sendo assim possível atribuir semântica específica a esses objetos através da descrição de seus atributos e manter a independência do conhecimento relativo a cada nível. O nível da imagem descreve os objetos passíveis de serem extraídos por algoritmos de processamento de imagem (embora esses algoritmos não tenham sido foco de estudo neste trabalho). O nível visual descreve objetos que são foco da atenção visual, tais como seções, interstícios e contornos. O nível semântico descreve os objetos da aplicação capturados através de aquisição de conhecimento. A identidade dos objetos modelados é garantida através de relações de mapeamento entre cada dois níveis adjacentes. O domínio de aplicação deste trabalho foi a Petrografia Sedimentar, com o objetivo de extrair por inferência a qualidade em termos de porosidade e permeabilidade de rochas reservatório de petróleo. Com ajuda do especialista, foi modelado um método de solução de problemas para identificação do grau de compactação da rocha, que raciocina sobre os conhecimentos modelados utilizando a ontologia proposta. Foi implementado um sistema que permite a descrição dos objetos individualizados através da segmentação manual da imagem, mapeando os dados descritos para a ontologia e aplicando sobre ela o método de solução de problemas. Esse sistema gera como resultado o grau de compactação da rocha, cuja imagem foi assim descrita. Uma validação preliminar da abordagem foi realizada através da descrição de imagens de rochas fazendo uso do sistema desenvolvido, confrontando os resultados com os obtidos por um geólogo para as mesmas rochas observadas. Na metade das amostras descritas, o sistema atingiu o mesmo resultado do especialista e, na outra metade, obteve grande aproximação dos resultados. / In domains that have strong visual content, the image interpretation applying visual reasoning can be more effective in solving problems than the interpretation of pure textual or numeric data. However, the representation of visual knowledge is hard to be achieved since, most of time, we are dealing with implicit knowledge for the observer. The representation ontologies allow the creation of structures for assisting the capture of this kind of knowledge, in order to associate a symbolic representation and semantic meaning to what it being visualize. The formalization of the visual knowledge allows its application for inference process, resulting in the automatic interpretation of image. The goal of this work is the definition of ontological constructs that allow describing the visual aspects presented in an image, giving more emphasis in the evidences captured by visual attention than in the physical aspects of the objects. These aspects are associated to the physical objects as well as to the objects described in the domain knowledge level. Separate representation ontologies were defined for each level, making possible to associate specific semantic content to the objects through the description of the attributes and to keep the independence of the knowledge related to each level. The image level describes the objects that are possible of being extracted by image processing algorithms (although these algorithms were not studied in this work). The visual knowledge describes the objects that capture the visual attention, such as sections, interstices and borders. The semantic level describes the application objects elicited by knowledge acquisition methods. The identity of the modeled objects is guaranteed through the mapping relation defined between each two adjacent levels. The application domain of this work is the Sedimentary Petrography, with the goal of extracting by inference methods the porosity and permeability quality of petroleum reservoir-rocks. With the aid of the expert, a problem-solving method that reasons over the knowledge formalized through the proposed ontology was modeled for the identification of the compaction level of the rock. Furthermore, it was implemented a system that supports the description of the objects individualized through a manual segmentation of the image. The described data was mapped to the ontology and the problem-solving method was applied to define the level of compaction. A preliminary validation was developed comparing the results achieved by the system with the manual interpretation done by the expert with the same rock samples. With the half of the described samples the system achieved the same results of the expert and has got strong approximation in the other half.