京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第18617号 / 情博第541号 / 新制||情||96(附属図書館) / 31517 / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 田中 克己, 教授 吉川 正俊, 教授 黒橋 禎夫 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

<p>The purpose of this study is to present and analyze the translation of seven selected concessive conjuncts – anyway, however, although, though, still, nonetheless and yet – in Dan Brown’s novel Angels and Demons translated by Ola Klingberg, by means of a comparative method combined with a qualitative analysis. Background and theory are mainly based on Altenberg (1999, 2002) for the conjuncts and Ingo (1991) for translation strategies. The aim is fulfilled by answering the three research questions: 1. How does Klingberg translate the seven selected concessive conjuncts into Swedish? 2. What factors influence the choice of translation alternative? 3. What kinds of strategies does Klingberg use? The main result is that the conjuncts translate into many different alternatives, although most frequently into the Swedish adversative men, followed by a Swedish concessive like ändå. However, the analysis of anyway is inconclusive because there were not enough tokens. The main conclusion is that translation is a difficult area to be involved in since numerous aspects affect the choice of translation alternative, even though it is shown that it is definitely possible to translate more or less ‘correctly’. A second conclusion is that some words are more likely to be translated with a particular word than others.</p>

The purpose of this study is to present and analyze the translation of seven selected concessive conjuncts – anyway, however, although, though, still, nonetheless and yet – in Dan Brown’s novel Angels and Demons translated by Ola Klingberg, by means of a comparative method combined with a qualitative analysis. Background and theory are mainly based on Altenberg (1999, 2002) for the conjuncts and Ingo (1991) for translation strategies. The aim is fulfilled by answering the three research questions: 1. How does Klingberg translate the seven selected concessive conjuncts into Swedish? 2. What factors influence the choice of translation alternative? 3. What kinds of strategies does Klingberg use? The main result is that the conjuncts translate into many different alternatives, although most frequently into the Swedish adversative men, followed by a Swedish concessive like ändå. However, the analysis of anyway is inconclusive because there were not enough tokens. The main conclusion is that translation is a difficult area to be involved in since numerous aspects affect the choice of translation alternative, even though it is shown that it is definitely possible to translate more or less ‘correctly’. A second conclusion is that some words are more likely to be translated with a particular word than others.

Surprise is an important component of serendipity. In this research, we address the problem of measuring the capacity of a recommender system at embedding surprise in its recommendations. We show that changes in surprise of an item owing to the growth in user experience, as well as to the increase in the number of items in the repository, are not taken into account by the current metrics and evaluation methods. As a result, in so far as the time elapsed between two measurements grows, they become increasingly incommensurable. This poses as an additional challenge in the assessment of the degree to which a recommender is exposed to unfavourable conditions, such as over-specialisation or filter bubble. We argue that a) surprise is a finite resource in any recommender system, b) there are limits to the amount of surprise that can be embedded in a recommendation, and c) these limits allow us to create a scale up in which two measurements that were taken at different moments can be directly compared. By adopting these ideas as premises, we applied the deductive method to define the concepts of maximum and minimum potential surprises and designed a surprise metric called \"normalised surprise\" that employs these limits. Our main contribution is an evaluation method that estimates the normalised surprise of a system. Four experiments were conducted to test the proposed metrics. The aim of the first and the second experiments was to validate the quality of the estimates of minimum and maximum potential surprise values obtained by means of a greedy algorithm. The first experiment employed a synthetic dataset to explore the limits to surprise to a user, and the second one employed the Movielens-1M to explore the limits to surprise that can be embedded in a recommendation list. The third experiment also employed the Movielens-1M dataset and was designed to investigate the effect that changes in item representation and item comparison exert on surprise. Finally, the fourth experiment compares the proposed and the current state-of-the-art evaluation method in terms of their results and execution times. The results obtained from the experiments a) confirm that the quality of the estimates of potential surprise are adequate for the purpose of evaluating normalised surprise; b) show that the item representation and comparison model that is adopted has a strong effect on surprise; and c) indicate an association between high degrees of surprise and negatively skewed pairwise distance distributions, and also indicate a significant difference in the average normalised surprise of recommendations produced by a factorisation algorithm when the surprise employs the cosine or the Euclidean distance / A surpresa é um componente importante da serendipidade. Nesta pesquisa, abordamos o problema de medir a capacidade de um sistema de recomendação de incorporar surpresa em suas recomendações. Mostramos que as mudanças na surpresa de um item, devidas ao crescimento da experiência do usuário e ao aumento do número de itens no repositório, não são consideradas pelas métricas e métodos de avaliação atuais. Como resultado, na medida em que aumenta o tempo decorrido entre duas medições, essas se tornam cada vez mais incomensuráveis. Isso se apresenta como um desafio adicional na avaliação do grau em que um sistema de recomendação está exposto a condições desfavoráveis como superespecialização ou filtro invisível. Argumentamos que a) surpresa é um recurso finito em qualquer sistema de recomendação; b) há limites para a quantidade de surpresa que pode ser incorporada em uma recomendação; e c) esses limites nos permitem criar uma escala na qual duas medições que foram tomadas em momentos diferentes podem ser comparadas diretamente. Ao adotar essas ideias como premissas, aplicamos o método dedutivo para definir os conceitos de surpresa potencial máxima e mínima e projetar uma métrica denominada \"surpresa normalizada\", que emprega esses limites. Nossa principal contribuição é um método de avaliação que estima a surpresa normalizada de um sistema. Quatro experimentos foram realizados para testar as métricas propostas. O objetivo do primeiro e do segundo experimentos foi validar a qualidade das estimativas de surpresa potencial mínima e máxima obtidas por meio de um algoritmo guloso. O primeiro experimento empregou um conjunto de dados sintético para explorar os limites de surpresa para um usuário, e o segundo empregou o Movielens-1M para explorar os limites da surpresa que pode ser incorporada em uma lista de recomendações. O terceiro experimento também empregou o conjunto de dados Movielens-1M e foi desenvolvido para investigar o efeito que mudanças na representação de itens e na comparação de itens exercem sobre a surpresa. Finalmente, o quarto experimento compara os métodos de avaliação atual e proposto em termos de seus resultados e tempos de execução. Os resultados que foram obtidos dos experimentos a) confirmam que a qualidade das estimativas de surpresa potencial são adequadas para o propósito de avaliar surpresa normalizada; b) mostram que o modelo de representação e comparação de itens adotado exerce um forte efeito sobre a surpresa; e c) apontam uma associação entre graus de surpresa elevados e distribuições assimétricas negativas de distâncias, e também apontam diferenças significativas na surpresa normalizada média de recomendações produzidas por um algoritmo de fatoração quando a surpresa emprega a distância do cosseno ou a distância Euclidiana

Une part importante de l’activité de communication humaine est dédiée au récit d’événements (fictifs ou non). Ces récits doivent être cohérents et intéressants pour être pertinents. Dans le domaine de la génération automatique de récits, la question de l’intérêt a souvent été négligée, ou traitée via l’utilisation de méthodes ad hoc, au profit de la cohérence des structures narratives produites. Nous proposons d’aborder le processus de création des récits sous l’angle de la modélisation quantitative de critères de pertinence narrative via l’application d’un modèle cognitif de l’intérêt événementiel. Nous montrerons que cet effort de modélisation peut servir de guide pour concevoir un modèle cognitivement plausible de génération de narrations. / Humans devote a considerable amount of time to producing narratives. Whatever a story is used for (whether to entertain or to teach), it must be relevant. Relevant stories must be believable and interesting. The field of computational generation of narratives has explored many ways of generating narratives, especially well-formed and understandable ones. The question of what makes a story interesting has however been largely ignored or barely addressed. Only some specific aspects of narrative interest have been considered. No general theoretical framework that would serve as guidance for the generation of interesting and believable narratives has been provided. The aim of this thesis is to introduce a cognitive model of situational interest and use it to offer formal criteria to decide to what extent a story is relevant. Such criteria could guide the development of a cognitively plausible model of story generation.