Feature selection is one of the important data preprocessing steps in data mining. The feature selection problem involves finding a feature subset such that a classification model built only with this subset would have better predictive accuracy than model built with a complete set of features. In this study, we propose two hybrid methods for feature selection. The best features are selected through either the hybrid methods or existing feature selection methods. Next, the reduced dataset is used to build classification models using five classifiers. The classification accuracy was evaluated in terms of the area under the Receiver Operating Characteristic (ROC) curve (AUC) performance metric. The proposed methods have been shown empirically to improve the performance of existing feature selection methods.

The need for navigation in hypertext requires cognitive resources, and this can result in disorientation and cognitive overhead. So, information retrieval and text comprehension are impaired. It is not possible to construct a coherent mental representation of text content (situation model), which is essential for text comprehension. The question is, how can, based on hypertext studies and knowledge about text comprehension, suggestions for hypertext design be found. In this study, the influence of different navigation possibilities and linking structures of hypertext on orientation problems, text comprehension, and efficiency of information retrieval was investigated. First, linear text and hierarchic structured hypertext with a graphical overview over the text structure were compared. Furthermore, text comprehension processes were focused more intensively. Therefore, the influence of the coherence of linking structure and of working memory load on interaction with the text, text comprehension, and efficiency of information retrieval was investigated. Results show that disadvantages of hypertext concerning orientation problems can be compensated with the aid of a graphical overview which is usable for navigation. This orientation and navigation aid is also an advantage for the speed of information retrieval. In contrast, for text comprehension coherence of the linking of individual text nodes plays an essential role. Only if hypertext is constructed in a way that a coherent reading sequence is suggested to the reader, the user is able to construct a coherent mental representation about the text content. It becomes apparent that different tasks, in this case reading a text vs. information retrieval, make different demands to hypertext. To some extend, the results were only shown with high working memory load, which shows the influence of cognitive resources. / Die Notwendigkeit der Navigation in Hypertexten beansprucht kognitive Ressourcen und führt leicht zu einer Desorientierung und einer kognitiven Überlastung. Sie erschwert damit das Auffinden von Informationen und beeinträchtigt das Verstehen des Textes, d.h. es kann keine kohärente mentale Repräsentation über den Textinhalt (Situationsmodell) aufgebaut werden, was jedoch in der Textverstehensforschung als zentraler Punkt des Verstehens betrachtet wird. Die Frage ist nun, wie aus den bisherigen Erkenntnissen zu Hypertexten und aus dem Wissen über Textverstehensprozesse Hinweise für eine optimale Hypertextgestaltung abgeleitet werden können. In der vorliegenden Arbeit wurde dazu der Einfluss verschiedener Navigationsmöglichkeiten und der Verknüpfungsstruktur von Hypertexten auf die Probleme des Nutzers beim Umgang mit dem Hypertext und auf das Textverstehen bzw. den Wissenserwerb und die Effizienz der Informationssuche untersucht. Zunächst wurde dazu ein Vergleich von linear verknüpftem Text und einem hierarchisch strukturierten Hypertext mit einer graphischen Übersicht über die Textstruktur vorgenommen. Weiterhin wurden verstärkt die Textverstehensprozesse beim Umgang mit Hypertext betrachtet. Dazu wurde der Einfluss des Kohärenzgrades der Verknüpfung des Textes und der Arbeitsgedächtnisbelastung auf den Umgang mit dem Text, das Textverstehen und die Effizienz der Informationssuche untersucht. Die Ergebnisse zeigen, dass die Nachteile des Hypertextes hinsichtlich der Orientierungsprobleme durch eine navigierbare graphische Übersicht über die Textstruktur kompensiert werden können. Diese Strukturierungs- und Navigationshilfe erweist sich auch als Vorteil für die Schnelligkeit der Informationssuche. Für das Textverstehen hingegen spielt die Kohärenz der Verknüpfung der einzelnen Textknoten eine zentrale Rolle. Nur wenn der Hypertext so strukturiert ist, dass dem Nutzer eine zeitlich kohärente Leseweise nahegelegt wird, ist der Nutzer auch in der Lage, eine kohärente mentale Repräsentation über den Textinhalt aufzubauen. Es zeigt sich deutlich, dass unterschiedliche Aufgaben, hier Lesen eines Textes vs. Suchen nach Informationen, unterschiedliche Anforderungen an Hypertexte stellen. Teilweise werden die gezeigten Ergebnisse erst unter einer erhöhten Arbeitsgedächtnisbelastung deutlich, was den Einfluss kognitiver Ressourcen deutlich macht.

Thèse numérisée par la Division de la gestion de documents et des archives de l'Université de Montréal

Thesis (PhD) -- Macquarie University, Division of Information and Communication Sciences, Department of Computing, 2007. / "Submitted January 27 2007, revised July 27 2007". / Bibliography: p. 281-304. / Mode of access: World Wide Web. / xiv, 487 p., bound ill. (some col.)

This research was aimed at determining the ethno-ornithological knowledge of three Mushere villages close to the Dulu Forest, Nigeria to determine whether this might hold potential for developing a local conservation programme to protect the forest. The conservation objective of the research was aimed at determining the drivers of forest degradation, and possible means of mitigation. Using a mixed method approach, quantitative/qualitative data were collected in two years from different demographics (men, women, children, age differentials, occupation, urbanization). Oral interviews, semi-structured interviews, picture elicitation tasks, free-listing exercises and focus group discussions were employed in the data collection process. The findings revealed a relatively low ethno-ornithological knowledge, and a general indifference (ornitho-apatheia) towards birds. Knowledge transmission was predominantly through oral means while TEK acquisition was mainly through vertical and horizontal methods. While adults perceived birds as not valuable, children generally perceived birds as valuable. Cultural utilization and ecological salience were the main drivers of bird naming and knowledge. However, there was a high valuing of the Dulu forest, with the main drivers of the forest degradation being timber extraction. Overall, I concluded that the indifference of the Mushere towards birds revealed a lack of cultural appreciation of birds, leading to little TEK of birds, insufficient to encourage the use of TEK of birds in the conservation of the Dulu forest, and the use of birds as flagship domain for promoting conservation. However, a sustained approach towards encouraging birding activities could improve the perception of birds. I therefore suggest using an ecosystem approach in the conservation of the Dulu forest. Engaging the locals in dialogue, establishing a leadership structure for the management of the Dulu forest, providing alternative means of livelihoods are suggested as ways of mitigating the degradation of the Dulu forest.

Cette thèse aborde la problématique de l’acquisition des connaissances dans le cadre de la conception des Environnements Informatiques pour l’Apprentissage Humain (EIAH). Dans le contexte spécifique de ce travail, nous nous intéressons à des EIAH destinés à enseigner des méthodes de résolution de problèmes. De telles méthodes permettent, dans un domaine donné, de reconnaître la classe d’un problème et d’être capable de savoir quelle technique de résolution appliquer pour le résoudre. Le coût de conception de ces EIAH est cependant très élevé, en particulier du fait de l’élicitation des connaissances, qui nécessite non seulement une expertise dans le domaine concerné mais également en programmation.Afin de réduire le coût de conception de ces EIAH, et permettre à des auteurs (enseignants ou pédagogues plus experts) de pouvoir éliciter sans programmer les connaissances nécessaires à l’EIAH, nous proposons un processus d’acquisition interactive de ces connaissances. Ce processus est mis en œuvre à travers la conception d’un outil auteur : AMBRE-KB. Pour ce faire, nous avons d’abord proposé des méta-modèles qui permettent de décrire la forme des connaissances à acquérir. Ces connaissances ne sont pas celles d’un expert dans un domaine, mais les connaissances telles qu’on voudrait qu’elles fonctionnent chez l’apprenant à l’issue de l’apprentissage. En s’appuyant sur ces méta-modèles, nous avons ensuite proposé un processus d’acquisition de ces connaissances qui permet d’assister l’auteur lors de l’élicitation des connaissances nécessaires, en lui permettant de construire un modèle de connaissances spécifique à un domaine.Nous avons mené deux expérimentations pour évaluer le processus d’acquisition des connaissances et sa mise en œuvre dans l’outil AMBRE-KB. La première porte sur la complétude. Il s’agit de vérifier, pour un domaine donné, si les modèles de connaissances générés par AMBRE-KB permettent à l’EIAH de résoudre des problèmes. L’objectif de la deuxième expérimentation est de mesurer l’utilité et l’utilisabilité de AMBRE-KB. Les résultats des deux expérimentations sont satisfaisants. / The general issue we addressed in this thesis is the challenge of knowledge elicitation in Intelligent Tutoring Systems (ITS). In the context of this work, we are interested in ITS teaching problem solving methods. Teaching methods for solving problems consists in teaching students how to think about the problem before starting its resolution. In a given field, such a method is based on a categorization of problems. Knowing to recognize the class of a problem enables students to choose the resolution technique associated with this class. However, designing such ITS is tedious and costly, and specially require expertise in the application domain and in programming.In order to reduce the design cost of these ITS and to enable an author (for example a teacher) to be able to elicit knowledge needed without programming, we propose an interactive knowledge elicitation process. This process is implemented through the design of an authoring tool: AMBRE-KB. For that, we first propose meta-models for the knowledge to be acquired. This knowledge is not an expert knowledge, but knowledge such as we would want that they work at the end of the learning. Next, we propose a knowledge acquisition process based on these meta-models, which enable the author to be assisted in the elicitation process enabling him/her to build specific knowledge models for a given domain.We conducted two experiments to evaluate the knowledge acquisition process and its implementation in the AMBRE-KB tool. The first relates to completeness. The aim is to verify, for a given domain, whether the knowledge models generated by AMBRE-KB enable the solver to solve problems. The aim of the second experiment is to measure the utility and usability of AMBRE-KB. The results of both experiments are satisfactory.

Knowledge is now considered the most important organizational resource, surpassing other resources like land and capital. It has, therefore, been acknowledged that knowledge can play an important role in ensuring an organization’s competitive edge. The purpose of this study was to investigate if knowledge sharing is being used to improve service delivery at the National University of Lesotho’s Thomas Mofolo Library. The researcher held the view that Librarians at Thomas Mofolo Library have different sets of skills which, if combined, could improve service delivery. By not sharing and retaining this existing wealth of knowledge, the researcher claimed that when librarians retire or resign from work, they will certainly take with them the knowledge they possess and the result of this knowledge loss is that the Library may be plagued by an inability to learn from the past experiences, which leads to reinvented wheels, unlearned lessons and the pattern of repeated mistakes. Both qualitative and quantitative methods were employed in the case study design in order to allow for multiple methods of data collection. Data were collected by means of questionnaires and interviews. Questionnaires were administered to all librarians who were available at the time and purposive sampling was used to determine interview participants. Out of the 25 questionnaires administered, 15 were returned, providing a response rate of 60%. The data collected by means of questionnaires was processed using Microsoft Access and analyzed using the Statistical Package for Social Science (SPSS) software (Version 17). The results of analysis were exported into Microsoft Excel for visual presentation and reporting of the results. The data from the interview sessions was analyzed manually by content analysis, using the notes that were taken by the researcher from the respondents during the interview sessions. The findings pointed to the fact that knowledge sharing does occur at TML, although mostly in an informal manner. This was largely due to a number of impediments such as lack of trust and the absence of motivations and rewards. The study concluded by recommending a number of initiatives that could be implemented in order to retain knowledge within the Library. The recommendations included developing a knowledge management strategy and formalizing knowledge sharing by formulating the desired policies. / Information Science / M.A. (Information Science)

Os avanços tecnológicos da última década tem feito dos computadores um elemento de contribuição essencial para os processos de solução de problemas e de tomada de decisão cooperativos. Hoje, alem do interesse mantido nos sistemas de solução de problemas, cujo raciocínio a baseado no processo de decisão de um único individuo (conhecidos por SE's), o esforço das pesquisas, em Inteligência Artificial, esta centrado no sentido de conceber sistemas que permitam a interação cooperativa entre diversos indivíduos participantes do processo, sejam esses humanos ou sistemas computacionais. A solução de problemas cooperativos, dentro do escopo geral da Inteligência Artificial (IA), é assunto analisado sob dois aspectos diferentes. O primeiro, mais antigo, identifica, como agentes de um dialogo, o sistema computacional e o seu usuário, onde pesquisas estão centradas no estudo da interação homem-máquina. Os esforços desta área de pesquisa tem sido no sentido de conceber, aos sistemas, capacidades de comunicação muito mais ricas do que aquelas oferecidas por sistemas de solução de problemas tradicionais, isto é, permitir aos sistemas compartilhar a solução de um problema, tomando o usuário um agente muito mais ativo e participativo. O segundo aspecto situa-se na área de Inteligência Artificial Distribuída (IAD), uma nova concepção de IA que acompanha o avanço da tecnologia de desenvolvimento de maquinas paralelas e a difusão, em larga escala, de sistemas computacionais distribuídos. Seus esforços são no sentido de conceber sistemas compostos de múltiplos sub-sistemas, capazes de resolver problemas complexos autonomamente. de forma cooperativa. Este trabalho se insere no contexto da interação homem-máquina. São apresentados métodos e estratégias para o fornecimento de capacidades cooperativas ao sistema. A descrição de uma arquitetura para Sistemas Especialistas (SE), baseada em raciocínio meta-nível, é apresentada com o intuito de enriquecer as capacidades de explanação e aquisição de conhecimentos desses sistemas. Consideramos que as ferramentas de explanação e aquisição de conhecimentos são fundamentais para a construção de diálogos cooperativos entre o sistema e o usuário. A ferramenta de explanação é o componente do SE responsável pela geração de justificativas sobre as conclusões do sistema. Ela permite ao sistema tornar explicito o seu raciocínio, fornecendo capacidades de argumentação sobre a validade de suas conclusões. O sistema, através desta ferramenta, tem condições de explicar suas ações, conclusões. escolhas e perguntas feitas ao usuário, permitindo, dessa forma, ao próprio usuário. através de um diálogo cooperativo. comparar seus conhecimentos e estratégias. concordando ou discordando do sistema. A ferramenta de aquisição de conhecimento. outro modulo importante num processo cooperativo, permite ao sistema aprender incrementalmente, através da aquisição de novos conhecimentos, bem como da reestruturação de alguns conhecimentos ou regras com falhas. Analisa-se um problema real, cuja solução e concebida através da interação homem-máquina. embora, intuitivamente, seja apresentada uma abordagem multi-agente para o problema, no final deste trabalho, com o intuito de apontar a evolução que terá essa pesquisa. Como produto deste trabalho de pesquisa, desenvolveu-se, dentro do projeto Inteligência Artificial Distribuída do grupo de Inteligência Artificial do CPGCC da UFRGS. urn sistema denominado SETA. O sistema permite a criação de SE's dedicados a auxiliar o medico na prescrição farmacológica de qualquer grupo de patologia clinica. A representação do conhecimento aplicado foi desenvolvida com o intuito de facilitar a atividade de formulação de prescrições onde o conhecimento esta estruturado em níveis de representação que denotam os conhecimentos clínico e farmacológico separadamente. Cada SE, desenvolvido pelo SETA, permite oferecer justificativas claras ao usuário sobre a prescrição farmacológica indicada pelo sistema, através de explanações do tipo how. why e why not. Oferece, ainda, facilidades de aquisição de conhecimento, permitindo a modificação do conhecimento do sistema através de um modulo interativo, cuja interface foi construída no sentido de permitir uma comunicacc7o natural entre os agentes. ou seja. o sistema e o medico especialista. Resumindo, a interação cooperativa homem-máquina é concebida através das facilidades de explanação e aquisição de conhecimento, levando a incorporação explicita de meta-conhecimento ao sistema. / The technological advance of the past decade turned computers into an element of essential contribution for the cooperative problem-solving and decision-taking processes. Today, besides the interest kept in problem-solving systems, whose reasoning is based upon the process of decision of a single individual (known as ES), the effort of researchs in Artificial Intelligence is to create systems that allow cooperative interaction among various individuals participating in the process, these being either human beings or computer systems. The solution of cooperative problems, within the general scope of Artificial Intelligence, is a subject analysed under two different aspects. The first one, out of date, identifies the computer system and its user as agents in a dialogue, and the researches are concentrated on the study of man-machine interaction. The efforts in this area of research have been to grant the systems communication abilities much richer than those offered by tradicional problem-solving systems. that is, that allow the systems to share the solution of a problem. causing the user to be more active and participating. The second aspect is located in the area of Distributed Artificial Intelligence (DAI), a new conception of AI that goes with the improvement of the technology of development of parallel machines and the diffusion on a large scale of distributed computer systems. Efforts have been made to create systems made up of multiple sub-systems capable of solving complex problems by themselves, in a cooperative way. This work is inserted in the context of man-machine interaction. It presents methods and strategies to supply the system with cooperative abilities. The description of an architecture for Expert Systems (ES), based upon meta-level reasoning, is presented with the purpose of improving the abilities of explanation and knowledge acquisition of these systems. We consider explanation tools and knowledge acquisition to be fundamental to the construction of cooperative dialogues between the system and the user. The explanation tool is the ES component responsible for the generation of justifications about the system conclusions. It allows the system to make its reasoning explicit, providing it with arguing abilities about the effectiveness of its conclusions. The system. through this tool, is able to explain its actions, conclusions, choices and questions put to the user, thus allowing the user, through a cooperative dialogue, to compare his knowledge and strategies, to agree or disagree with the system. The knowledge acquisition tool, another important unit in a cooperative process, allows the system to learn more and more through the acquisition of new knowledge as well as through the restructuration of knowledge or rules that have failed. A real problem is analysed here and its solution is conceived through manmachine interaction. We also present, at the end of this work, a multi-agent approach for the problem, in order to show how this research will evolve. This research work resulted in the development, within the Distributed Artificial Intelligence project of the Artificial Intelligence group of the CPGCC of UFRGS, of a system called SETA. This system permits the creation of ES dedicated to help doctors prescribe medicines for any groups of clinical pathology. The knowledge representation used was developed with a view to facilitate the making of prescriptions, and the knowledge is organized in levels of representation that express clinical knowledge and pharmacological knowledge separately. Each ES developed by SETA can offer reasonable justifications to users about the pharmacological prescription indicated by the system through explanations such as how, why and why not. It is also ready to acquire knowledge. allowing the system to alter knowledge through an interactive unit whose interface was built to permit a natural communication between the agents. that is, the system and the medical specialist. In short, man-machine cooperative interaction is based upon readiness for explanation and knowledge acquisition, leading to an explicit assimilation of meta-knowledge by the system.

Os avanços tecnológicos da última década tem feito dos computadores um elemento de contribuição essencial para os processos de solução de problemas e de tomada de decisão cooperativos. Hoje, alem do interesse mantido nos sistemas de solução de problemas, cujo raciocínio a baseado no processo de decisão de um único individuo (conhecidos por SE's), o esforço das pesquisas, em Inteligência Artificial, esta centrado no sentido de conceber sistemas que permitam a interação cooperativa entre diversos indivíduos participantes do processo, sejam esses humanos ou sistemas computacionais. A solução de problemas cooperativos, dentro do escopo geral da Inteligência Artificial (IA), é assunto analisado sob dois aspectos diferentes. O primeiro, mais antigo, identifica, como agentes de um dialogo, o sistema computacional e o seu usuário, onde pesquisas estão centradas no estudo da interação homem-máquina. Os esforços desta área de pesquisa tem sido no sentido de conceber, aos sistemas, capacidades de comunicação muito mais ricas do que aquelas oferecidas por sistemas de solução de problemas tradicionais, isto é, permitir aos sistemas compartilhar a solução de um problema, tomando o usuário um agente muito mais ativo e participativo. O segundo aspecto situa-se na área de Inteligência Artificial Distribuída (IAD), uma nova concepção de IA que acompanha o avanço da tecnologia de desenvolvimento de maquinas paralelas e a difusão, em larga escala, de sistemas computacionais distribuídos. Seus esforços são no sentido de conceber sistemas compostos de múltiplos sub-sistemas, capazes de resolver problemas complexos autonomamente. de forma cooperativa. Este trabalho se insere no contexto da interação homem-máquina. São apresentados métodos e estratégias para o fornecimento de capacidades cooperativas ao sistema. A descrição de uma arquitetura para Sistemas Especialistas (SE), baseada em raciocínio meta-nível, é apresentada com o intuito de enriquecer as capacidades de explanação e aquisição de conhecimentos desses sistemas. Consideramos que as ferramentas de explanação e aquisição de conhecimentos são fundamentais para a construção de diálogos cooperativos entre o sistema e o usuário. A ferramenta de explanação é o componente do SE responsável pela geração de justificativas sobre as conclusões do sistema. Ela permite ao sistema tornar explicito o seu raciocínio, fornecendo capacidades de argumentação sobre a validade de suas conclusões. O sistema, através desta ferramenta, tem condições de explicar suas ações, conclusões. escolhas e perguntas feitas ao usuário, permitindo, dessa forma, ao próprio usuário. através de um diálogo cooperativo. comparar seus conhecimentos e estratégias. concordando ou discordando do sistema. A ferramenta de aquisição de conhecimento. outro modulo importante num processo cooperativo, permite ao sistema aprender incrementalmente, através da aquisição de novos conhecimentos, bem como da reestruturação de alguns conhecimentos ou regras com falhas. Analisa-se um problema real, cuja solução e concebida através da interação homem-máquina. embora, intuitivamente, seja apresentada uma abordagem multi-agente para o problema, no final deste trabalho, com o intuito de apontar a evolução que terá essa pesquisa. Como produto deste trabalho de pesquisa, desenvolveu-se, dentro do projeto Inteligência Artificial Distribuída do grupo de Inteligência Artificial do CPGCC da UFRGS. urn sistema denominado SETA. O sistema permite a criação de SE's dedicados a auxiliar o medico na prescrição farmacológica de qualquer grupo de patologia clinica. A representação do conhecimento aplicado foi desenvolvida com o intuito de facilitar a atividade de formulação de prescrições onde o conhecimento esta estruturado em níveis de representação que denotam os conhecimentos clínico e farmacológico separadamente. Cada SE, desenvolvido pelo SETA, permite oferecer justificativas claras ao usuário sobre a prescrição farmacológica indicada pelo sistema, através de explanações do tipo how. why e why not. Oferece, ainda, facilidades de aquisição de conhecimento, permitindo a modificação do conhecimento do sistema através de um modulo interativo, cuja interface foi construída no sentido de permitir uma comunicacc7o natural entre os agentes. ou seja. o sistema e o medico especialista. Resumindo, a interação cooperativa homem-máquina é concebida através das facilidades de explanação e aquisição de conhecimento, levando a incorporação explicita de meta-conhecimento ao sistema. / The technological advance of the past decade turned computers into an element of essential contribution for the cooperative problem-solving and decision-taking processes. Today, besides the interest kept in problem-solving systems, whose reasoning is based upon the process of decision of a single individual (known as ES), the effort of researchs in Artificial Intelligence is to create systems that allow cooperative interaction among various individuals participating in the process, these being either human beings or computer systems. The solution of cooperative problems, within the general scope of Artificial Intelligence, is a subject analysed under two different aspects. The first one, out of date, identifies the computer system and its user as agents in a dialogue, and the researches are concentrated on the study of man-machine interaction. The efforts in this area of research have been to grant the systems communication abilities much richer than those offered by tradicional problem-solving systems. that is, that allow the systems to share the solution of a problem. causing the user to be more active and participating. The second aspect is located in the area of Distributed Artificial Intelligence (DAI), a new conception of AI that goes with the improvement of the technology of development of parallel machines and the diffusion on a large scale of distributed computer systems. Efforts have been made to create systems made up of multiple sub-systems capable of solving complex problems by themselves, in a cooperative way. This work is inserted in the context of man-machine interaction. It presents methods and strategies to supply the system with cooperative abilities. The description of an architecture for Expert Systems (ES), based upon meta-level reasoning, is presented with the purpose of improving the abilities of explanation and knowledge acquisition of these systems. We consider explanation tools and knowledge acquisition to be fundamental to the construction of cooperative dialogues between the system and the user. The explanation tool is the ES component responsible for the generation of justifications about the system conclusions. It allows the system to make its reasoning explicit, providing it with arguing abilities about the effectiveness of its conclusions. The system. through this tool, is able to explain its actions, conclusions, choices and questions put to the user, thus allowing the user, through a cooperative dialogue, to compare his knowledge and strategies, to agree or disagree with the system. The knowledge acquisition tool, another important unit in a cooperative process, allows the system to learn more and more through the acquisition of new knowledge as well as through the restructuration of knowledge or rules that have failed. A real problem is analysed here and its solution is conceived through manmachine interaction. We also present, at the end of this work, a multi-agent approach for the problem, in order to show how this research will evolve. This research work resulted in the development, within the Distributed Artificial Intelligence project of the Artificial Intelligence group of the CPGCC of UFRGS, of a system called SETA. This system permits the creation of ES dedicated to help doctors prescribe medicines for any groups of clinical pathology. The knowledge representation used was developed with a view to facilitate the making of prescriptions, and the knowledge is organized in levels of representation that express clinical knowledge and pharmacological knowledge separately. Each ES developed by SETA can offer reasonable justifications to users about the pharmacological prescription indicated by the system through explanations such as how, why and why not. It is also ready to acquire knowledge. allowing the system to alter knowledge through an interactive unit whose interface was built to permit a natural communication between the agents. that is, the system and the medical specialist. In short, man-machine cooperative interaction is based upon readiness for explanation and knowledge acquisition, leading to an explicit assimilation of meta-knowledge by the system.

Cette thèse analyse les stratégies de traduction et l’acquisition de connaissances à partir d’entretiens et de questionnaires envoyés à des traducteurs techniques indépendants. L’objectif est de voir si les stratégies de traduction sont influencées par des facteurs que le traducteur peut plus ou moins maîtriser et si l’expérience, la formation en traduction et-ou dans un domaine et, les langues de travail jouent aussi un rôle. Il apparaît que la pression du temps influence la décision d’accepter ou de refuser une traduction et que sous cette pression, le traducteur ne consulte pas les mêmes documents, acquiert moins de connaissances et, passe moins de temps à l’archivage. Le manque de sources pose des problèmes aux traducteurs de langues peu répandues. La création terminologique, l’analyse du texte, l’aide de la source humaine et l’archivage des informations sont plus systématiques pour ces traducteurs que pour ceux de langues répandues. Le traducteur spécialisé attache moins d’importance à la technicité du texte de départ que celui sans spécialisation, achète plus de documents et évalue plus facilement le temps qu’il consacrera à la recherche documentaire. Les « jeunes » traducteurs voient un lien fort entre la technicité du texte et sa difficulté, préfèrent utiliser des sources donnant des réponses immédiates aux problèmes, consultent et achètent moins de sources sur support papier que les traducteurs expérimentés. Internet a marginalisé l’utilisation et les achats de sources sur support papier mais ne permet pas forcément de se spécialiser. Avec Internet, le client accorde des délais plus courts et les traducteurs acceptent de traduire des textes plus techniques. / This thesis analyzes translation strategies and knowledge acquisition. It is based on interviews and questionnaires from a sample of freelance technical translators. The purpose is to investigate if translation strategies are influenced by factors translators can handle more or less, such as experience, training and working languages. It appears that time pressure has an influence on the decision to accept or refuse a translation. When under pressure, translators do not use the same documents, get less knowledge and spend less time storing information. The lack of documents is problematic for translators whose languages are rarely spoken and read in their working environment. They create more terms in the target language and analyze the source text more in-depth than translators of widespread languages ; they also get help from the human source and store information more often than translators of widespread languages. Specialized translators give more importance to the technicity of the source text than non-specialized translators ; they buy more documents and are more able to assess the time they will take to do documentary research. “Young” translators think there is a strong link between the technicity of the source text and its level of difficulty. They also prefer to directly use documents matching what they look for, use and buy less paper documents than experienced translators. Translators use and buy less paper documents, and accept to translate more technical texts than during the pre-Internet period. Clients also give them shorter deadlines than before.