Knowledge Acquisition (KA) from text requires that a large quantity of prior knowledge be made available to the Natural Language Processing (NLP) system. This prior knowledge is called foundational knowledge. The question of where foundational knowledge comes from in the first place is one of the biggest problem facing NLP. Conventionally, foundational knowledge has been hand-crafted on a task- and domain-specific basis. However, it is difficult to determine beforehand exactly what knowledge will be required. It has been shown within the TANKA project that a potential solution to this problem is to use surface NLP. Surface NLP relies solely on syntax and on the help of a user to elicit knowledge from text, hence effectively eliminating the need for prior-hand crafting of foundational knowledge. However, the domain knowledge obtained in this manner from a text contains gaps. The work presented in this thesis consisted in finding a better method than prior hand-crafting to acquire the knowledge needed to fill those gaps. The method presented, called Lazy KA, uses examples (short NL stories) and failures of an explanation mechanism such as EBL to find these gaps and to interactively and incrementally learn the required new knowledge. When the explanation of a particular example fails, the user is guided through a process that leads to the acquisition of the missing knowledge. Initially, the user is heavily involved, but as more examples are processed, the user becomes less and less involved. The convergence hypothesis, that is that the user interventions would decrease as examples are processed, was verified experimentally by using the prototype system FOKAS implementing these ideas. (Abstract shortened by UMI.)

On entend par "correcticiel" tout logiciel de correction de la langue ecrite capable d'intervenir non seulement sur l'orthographe des mots, mais aussi sur d'autres elements comme la syntaxe, la ponctuation, voire meme le sens. Comme aide a la redaction, le correcticiel suscite depuis peu un interet croissant chez les pedagogues. C'est d'ailleurs ce qui m'a amene a en evaluer le potentiel dans le contexte particulier de la didactique du theme francais. Pour les besoins de la cause, j'ai eu recours au Correcteur 101, un correcticiel reconnu pour sa capacite d'analyse et son approche pedagogique. Dans un premier temps, j'examine les lacunes linguistiques des etudiants en theme francais a partir des fautes relevees dans un corpus de traductions faites par un groupe de ces etudiants. Puis, au terme d'un essai sur un echantillon representatif du meme corpus, j'evalue la performance du Correcteur 101 par rapport aux faiblesses constatees chez les etudiants. On verra que 101 a pu reperer pres du quart des fautes de langue laissees par les etudiants, mais qu'il s'est avere inefficace la ou ceux-ci en auraient eu le plus besoin, c'est-a-dire en syntaxe. Quoi qu'il en soit, ce resultat non negligeable montre bien la capacitie du logiciel a detecter les erreurs; reste a savoir si, dans la pratique, les etudiants seront capables de les corriger. Pour eclairoir cette question, j'ai fait un essai avec des sujets anglophones et observe les reactions de ceux-ci aux diverses interventions du correcticiel. Appeles a corriger un texte francais sans aide exterieure, puis a le recorriger avec l'aide du Correteur 101, des sujets anglophones ont pu ameliorer leur performance de 29% dans le second cas. A la lumiere des resultats obtenus, je conclus que le correcticiel peut rendre de grands services aux etudiants en theme francais. S'ils savent bien s'en servir, ils pourront non seulement ameliorer la qualite de leurs textes, mais aussi parfaire leur connaissance le la langue francaise.

Abstract not available.

No description available.

The thesis aims at discussing one of the most influential and dominant current discourses about the Americas and their identity: the concept of americanidad. Fully developed in Hispanic Latin America in the first half of the 20th century, the concept gained great popularity in Quebec in the 1980s and 1990s (americanite), and has started to attain a certain level of recognition in Brazilian academe (americanidade). However, in its journey the concept americanidad changed, undergoing significant metamorphosis, acquiring new meanings and nuances. The thesis attempts to show the differences, similarities and parallelisms among the sister concepts americanidad, americanite and americanidade. In our attempt to map the concept and its diverse uses throughout the continent, we will also try to understand an important omission: why the analogous concept (americanity) has not been developed in the English-speaking portion of North Americas?

Based on the positivist ideal that the original text as immutable (Kasirer, 2001: 339), a series of negative connotations has been attributed to translation as a product in the legal community. In addition, the discipline of Translation Studies has devoted little attention to the status of the translated legal text. Faced with this marginalization, the goal of this thesis is to determine the importance of the hermeneutic role (Cote, 1999: 409) of authenticated translations in the interpretative process of multilingual treaties and, in a more general perspective, to show the status of authenticated translated texts. In fact, these texts, which often constitute legal precedents for subsequent agreements, can attain an equal, if not superior, position with respect to the so-called "original" text. First, this thesis presents an analysis of general rules of interpretation in order to determine the importance of consulting different authenticated translations of a multilingual treaty in the interpretative process. Second, the historical evolution of the role of the translated text in the interpretative process will be outlined. The goal here is to determine whether any changes have taken place that improved the status of translation in this interpretative context. Finally, in an attempt to be objective, a quantitative analysis shows the limitations of the hermeneutic role played by translations in the context of disputes settled pursuant to chapter 11 of the North American Free Trade Agreement.

Over the past several years, there has been such an enormous increase in the demand for translation services that translators have been struggling to keep up with it. This situation makes the use of computer translation tools ever more attractive. But once these programs have been installed, how can translators decide which one to use? Or whether it is worth the trouble of using them at all? The objective of this thesis is to answer these questions by proposing and testing an approach to predict which translation method---machine translation, translation memories or human translation---will be the best for a given text type. In this case "best" means the method that allows a translator to produce a "client-ready" translation as quickly as possible. After a review of the characteristics that, according to various authors, determine that a text is more suitable for a particular translation method, three different text types were analyzed and translated (from English to French) using the three methods listed above in order to establish the correlations between the characteristics of the texts and the method that gave the best results for each text. This data was used to develop a checklist of guidelines and to assign values to their various elements (e.g., the optimal number of words per sentence, the maximum number of sentences in the passive voice). Two pilot tests were then performed to verify the validity of two sets of values for predicting the best method on two new text types. After these texts were translated and the predictions verified, the values were adjusted and a final test was performed. The results of the final test support the conclusion that the approach proposed in this thesis can indeed help text allocators and translators choose the best method to use.

Our study builds on the metalexicographical studies devoted to the ideological representations contained within dictionaries. While the myth of objectivity of the dictionary keeps prevailing in the general public, many researchers are striving to deconstruct it. Their studies have helped depicting the sometimes subjective nature of the dictionary, particularly when describing the lexicon related to certain topics such as sexuality, religion or politics. Therefore, as part of our study, we wished to confirm the assumption, commonly accepted by researchers in the field, according to which the dictionary conveys ideological representations and subjectivity. The primary objective of our study was to explore and compare the lexicographical treatment of three terms pertaining to politics: nationalism, nationalist and nation, in six standard monolingual dictionaries of three important languages throughout the world, i.e. German, English and French. The dictionaries chosen for this study are: the Deutsches Universalworterbuch (or Duden), Wahrig's Deutsches Worterbuch, the Canadian Oxford Dictionary, the Random House Webster's Dictionary of the English Language, Le Petit Robert and Le Petit Larousse. We have chosen words from the political field since it is considered by many researchers to be conducive of ideological representations. Moreover, the analyses deal with four microstructural components: definitions, examples, usage labels and cross-references. Our second objective was to establish, in the light of the studies carried out, whether the concepts of nation, nationalism and nationalist are addressed objectively in the dictionaries related to the corpus or, on the contrary, whether their treatment lets a certain ideology show through. At the completion of our study, we were indeed able to observe a certain subjectivity in the lexicographical treatment suggested for the concepts of nationalism, nationalist and nation in some of the dictionaries related to the corpus.

Traditionally, translators have been encouraged to translate only into their mother tongue; however, this premise is being increasingly challenged in the translation literature. This thesis explores the related issue of translators working not into a foreign language, but into a foreign dialect. In particular, the dual objective of this thesis is first to demonstrate that corpora and corpus analysis tools provide a viable means for investigating and describing dialects, and second to determine whether a specially designed monodialectal corpus can be a useful resource for helping translators to produce adequate target texts in a non-native dialect. To meet these goals, two different yet related experiments are conducted. The first consists of a comparison between the textbook descriptions of the bonaerense dialect used in the Buenos Aires region of Argentina and the actual linguistic features present in a corpus consisting of authentic texts written in this dialect. The results reveal disparity between the textbook descriptions and the corpus contents, proving that a corpus is a versatile resource that can be used to learn more about dialects. The second experiment consists of a comparison of two sets of translations into bonaerense prepared by a group of translators from Spain. The first set of translations is carried out using conventional non-corpus-based resources (e.g. term banks, dictionaries), while the second set is produced using the above-mentioned monodialectal corpus of bonaerense. The results of this second experiment provide evidence suggesting that a monodialectal corpus can indeed facilitate the production of a target text in a non-native dialect.

Translation Memory (TM) systems are currently among the most popular translation tools available on the market. Clients today are looking for productivity and quality, but at the same time are looking to reduce costs. A TM stores portions of previously translated texts in a database so that they can be reused in new texts. There are two ways of working with a TM: interactive mode and pre-translation mode. In interactive mode, the TM system offers proposals to the translator, who can choose to accept and adapt them, or refuse them and translate from scratch. In pre-translation mode, the TM system's proposals are automatically inserted into the new source text, producing a hybrid text containing a mixture of source and target language elements. The translator's job is then to verify the translation of the automatically replaced portions of text, as well as to translate any text that was left in the source language. As the demand for translation continues to rise, more and more translators are looking to TMs to help increase productivity; however, for a variety of reasons, such as cost and incompatible file formats, they do not always have access to a useful TM. Therefore, translators need access to the TM database of the client they are working for in order to complete their translations. However, clients are increasingly hesitant to give out their databases due to proprietary issues. For this reason, many clients are turning to the pre-translation option so that they can get their translations done without having to give out their TM databases. To date, however, no one has yet studied whether the choice of pre-translation vs. interactive translation has an effect on productivity, quality and translator satisfaction. This thesis aims to go some way towards filling this gap by designing and testing a methodology to compare these two methods of working with a TM system. The thesis is divided into three main parts. Part I provides background information on TMs through a literature survey, the findings of which form the basis for three hypotheses concerning the ways in which interactive and pre-translation will impact translation productivity and quality, as well as translator satisfaction. Part II describes an experiment that was designed and conducted to investigate these hypotheses. Finally, Part III evaluates the research carried out in this thesis and suggests possible ways to expand it in the future.