When performing web searches, users manually open a web browser, direct it to a search engine, input keywords, and finally manually filter and select relevant results. This repetitive task can negatively impact the user's experience, something the automation and personalization of web search can address. This thesis presents PALTask, an Instant Messaging (IM) application that exploits context of both the user and their conversation in order to automate and personalize related web tasks such as web searches relevant to the conversation. PALTask dynamically gathers context and provides feedback from the user and the system at runtime including keywords from the conversation and running them through various search services such as YouTube and Google to retrieve relevant results. This thesis also explores various natural language processing (NLP) tasks such as keyword extraction, sentiment analysis, and stemming. These NLP tasks help in the collection of dynamic context at runtime, identifying personalized context, and analyzing it to improve the user's experience. We also present our keyword ranking algorithm which aims to improve accuracy when retrieving web resources. / Graduate

Session ID: TH-G3-5 / Joint 4th International Conference on Soft Computing and Intelligent Systems and 9th International Symposium on advanced Intelligent Systems, September 17-21, 2008, Nagoya University, Nagoya, Japan

Catchwords and catchphrases denoting crucial ideological concepts or disqualifying an opponent serve as indices pointing to the line of demarcation between political camps. Using the example of three catchwords (Patriotyzm genetyczny genetic patriotism, pka kulturowa cultural shelf, Palikotyzacja X-a Palikotization of X) emerging in the aggravated political debate in Poland between 2000 and 2009, and mainly signaling affiliation with the national conservative PiS-party, a twofold approach combining methods from discourse analysis and cognitive linguistics is proposed. On the one hand, the lexical items are analyzed as keywords in statu nascendi which are disputed, quoted or ridiculed and serve as intertextual hinges which still have the potential to evoke the communicative situation in which they have been created. On the other hand, the selected newly created lexical items are analyzed as conceptual integration networks within the framework of blending theory. The blending analysis reveals the underlying logic of the novel conceptual structure and displays the explanative and argumentative pattern suggested by the blend and condensed in the catchword. By tracing the disputes elicited by these catchwords and analyzing sequences in which these newly created concepts are contested or maintained from a cognitive perspective, we can see how the logic suggested by the blend is either perpetuated or perverted in the process of the ongoing dissemination of the catchword.

Les premières marqueteries de bois (la tarsia lignea) apparaissent en Italie dans les ateliers siennois du XIVe siècle avant de gagner l’ensemble des centres artistiques de la péninsule. Le Quattrocento voit le développement et la diffusion de cette technique décorative qui applique rigoureusement dans ses réalisations (vastes ensembles de choeurs et de sacristies, mobilier privé) la perspective et la théorie de la peinture, élaborées au même moment par les mathématiques. Les panneaux marquetés sont soulignés par des bordures dont le répertoire ornemental riche et varié est encore méconnu Leur très grande diversité témoigne d’une expérience approfondie des principes de composition et de variation des formes de base et de leur rationalisation, simultanément, cette recherche n’exclut pas le libre jeu de l’invention en vue de l’idéal de la varietas. Des premiers décors de la sagrestia delle Messe de Santa Maria del Fiore à Florence, jusqu’aux applications des placards de sacristie de Santa Maria di Monte Berico à Vicence, le répertoire ornemental des bordures évolue et s’enrichit Les réalisations des frères Canozi da Lendinara des années 1460 marque un moment significatif de leur renouvellement mais d’autres étapes et protagonistes, d’autres circuits de modèles et d’artistes-artisans restent à identifier. L’inventaire systématique des ornements des bordures marquetées doit permettre de restituer la grammaire et les principes stylistiques de formes jusqu’ici relativement négligées et dévaluées, leurs places et leurs fonctions dans l’art de la Renaissance épanouie. / Wood inlays (the tarsia lignea) were first made in Italy in the fourteenth century Sienese workshops before reaching all cultural centers of the peninsula. The Quattrocento witnesses the development and dissemination of this decorative technique, which applies to its creations and with much precision both perspective and theory of painting (large sets of choirs and sacristy, private furniture) developed during the same period in mathematics.The inlaid panels are highlighted by ornamental borders whose rich and varied repertory is still unknown.Their great diversity reflects an extensive experience of the principles of composition and variation of the basic forms and rationalization, simultaneously, this research does not exclude the free play of the invention to the ideal of varietas.

La corrida a motivé de nombreuses oeuvres d’artistes d’horizons et d’époques distincts, notamment depuis la période romantique. Aujourd’hui, la tauromachie est aussi bien associée à un sujet désuet, telle l’espagnolade, qu’à l’oeuvre majeure d’artistes aussi emblématiques que Goya ou Picasso. Mais la corrida est également la représentation de la violence, d’une pratique souvent perçue comme barbare, en même temps qu’une iconographie riche de sens par laquelle des artistes ont évoqué les maux de leur époque et tendu à l’homme le miroir de ses passions. Cet aspect naît avec l’oeuvre taurine de Goya – père de la représentation moderne de la tauromachie –, où le sujet devient grave, dramatique, dépouillé de ses détails anecdotiques, un sujet dont la portée devient parfois politique. C’est toute cette complexité, cette richesse et cette pluralité des regards portés sur la représentation artistique de la corrida que cette étude met en lumière. De la tauromachie en peinture : usages et imaginaires de Goya à Albarracín s’organise en trois temps. D’abord, l’étude de l’image de la corrida prise pour elle-même, ainsi que l’analyse des éléments qui en déterminent la réception. Ensuite, la mise en perspective historique des oeuvres dédiées à la tauromachie. Enfin, l’analyse des mythes auxquels l’image de la corrida peut être rattachée, et la définition de la poétique aux origines de nombre de ses représentations – où la corrida ne vise plus au local, mais à l’universel. L’appréhension de ses différentes strates et de la façon dont elles s’imbriquent, permet de comprendre l’intérêt des artistes pour la corrida. Dès lors, celle-ci acquiert dans la pratique artistique, par la poétique qui la meut, une dimension à la fois historique et universelle, lui faisant aborder les interrogations propres à la condition humaine. / The corrida has inspired numerous works of art by artists from different times and horizons, in particular from the romantic period onwards. Today tauromachy or bullfighting is both associated with out-of-date subjects such as the españolada (clichés about Spain or Spanish subjects) and major artworks by emblematic artists such as Goya or Picasso. But the corrida is also a representation of violence, a practice often considered to be barbarous ; at the same time it has been a rich iconography by which artists have portrayed the evils of their time and held out a mirror in which men could see their passions. This aspect began with Goya’s bullfighting works (Goya is considered the father of the modern representation of tauromachy). In the latter, the subject became serious, dramatic, stripped of its anecdotal details, a subject which sometimes had political ramifications. This study reveals the complexity, the richness and the plurality of representations of the art of bullfighting. Of tauromachy in painting : uses, customs and imaginary from Goya to Albarracin is divided up into three movements. Firstly, a study of the image of the corrida taken in and of itself, as well as an analysis of the elements which determine how it is perceived. Secondly, works dedicated to tauromachy are put into historical perspective. Lastly, an analysis of the myths the image of the corrida can be associated with, and the definition of the poetics at the origin of a number of these representations, in which the corrida no longer aims at the local but the universal. Understanding these different strata and the way they are involved in one another, allows us to understand why artists were interested in the art of bullfighting. From this point onwards, through art and the poetics which motivate these works, the corrida takes on both a historical and a universal dimension, which allows it to address questions belonging to the human condition.

Pas de résumé / No abstract

L’immigration des femmes originaires d’Afrique subsaharienne en direction de la France est fonction de la situation sociale, politique et économique du Continent africain. Depuis les années 80, elles sont de plus en plus nombreuses à émigrer alors même que la France remet progressivement en cause les conventions bilatérales qui dérogeaient à l’ordonnance n°45- 2685 du 2 novembre 1945 régissant en principe, l’entrée et le séjour des étrangers. Cela même en dehors du cadre du regroupement familial comme ce fut fréquemment le cas antérieurement. Les nombreuses modifications du régime général posé par l’ordonnance de 1945 résultent de la nécessaire adaptation de la législation française à des impératifs communautaires notamment depuis l’adoption du Traité de Maastricht en 1992. L’octroi de l’asile est une question qui intéresse les femmes notamment celles menacées d’excision, motif qui n’est admis que de manière restreinte par les autorités françaises contrairement à d’autres pays. Les candidates malheureuses à l’immigration sont contraintes de séjourner en France de manière illégale. Ces immigrées clandestines ne sont pas pour autant privées de tout droit. Cependant leur situation est parfois en contradiction avec les conventions relatives à la protection des droits de l’Homme. Ces femmes immigrées sont le plus souvent victimes de discrimination bien que leur niveau de compétence d’une génération à l’autre, soit en nette progression. En réaction aux problèmes d’intégration et désireuses de préserver leur culture, les communautés africaines perpétuent certaines coutumes qui sont en violation des conventions internationales portant sur les droits de l’Homme et de celles portant spécifiquement sur les droits de la Femme.Corrélativement ce repli communautaire rend plus difficile l’assimilation des immigrées qui souvent ne sont pas informées de leurs droits. / No abstract

Due to the unstructuredness and the lack of schema information of knowledge graphs, social networks and RDF graphs, keyword search has been proposed for querying such graphs/networks. Recently, various keyword search semantics have been designed. However, these keyword search semantics and algorithms encounter efficiency or scalability issues. In this thesis, we propose new three generic frameworks or index techniques to address these issues. The thesis results show that the keyword search on massive graphs under different scenarios can be effective and efficient, which would facilitate keyword search services on graphs in the real world. First, we study the keyword search on massive knowledge graphs. In particular, we propose a generic ontology- based indexing framework for keyword search, called Bisimulation of Generalized Graph Index (BiG-index), to enhance the search performance. The novelties of BiG-index reside in using an ontology graph GOnt to summarize and index a data graph G iteratively, to form a hierarchical index structure G. Regarding query evaluation, we transform a keyword search q into Q according to GOnt in runtime. The transformed query is searched on the summary graphs in G. The efficiency is due to the small sizes of the summary graphs and the early pruning of semantically irrelevant subgraphs. To illustrate BiG-index's applicability, we show popular indexing techniques for keyword search can be easily implemented on top of BiG-index. Our extensive experiments show that BiG-index clearly reduced the runtimes of popular keyword search algorithms. Second, we study the problem of keyword search on public-private graph. In many applications (e.g., social networks), users may prefer to hide parts or all of her/his data graphs (e.g., private friendships) from the public. This leads to a recent graph model, namely the public-private network model, in which each user has his/her own network. While there have been studies on public-private network analysis, keyword search on public-private networks has not yet been studied. Hence, we propose a new keyword search framework, called public-private keyword search (PPKWS). PPKWS consists of three major steps: partial evaluation, answer refinement, and answer completion. We select three representative ones and show that they can be implemented on the model with minor modifications. We propose indexes and optimizations for PPKWS. We have verified through experiments that, on average, the algorithms implemented on top of PPKWS run 113 times faster than the original algorithms directly running on the public network attached to the private network for retrieving answers that span through them. Third, we study the keyword search in distributed graph evaluation systems. In the recent research on query evaluation, parallel evaluation has attracted much interest. However, the study on keyword search on distributed graphs has still been limited. We propose a novel distributed keyword search framework called DKWS. We propose a notify-push paradigm which can exchange the upper bounds of answers across all the workers asynchronously. In particular, the workers notify the coordinator when the local upper bound is refined. The coordinator pushes the refined global upper bound to all the workers. Moreover, we propose an efficient and generic keyword search algorithm for the workers. We have implemented DKWS on top of GRAPE, a distributed graph process system from our previous research collaboration. Extensive experimental results show that DKWS outperforms current-state-of-art techniques

Due to the unstructuredness and the lack of schema information of knowledge graphs, social networks and RDF graphs, keyword search has been proposed for querying such graphs/networks. Recently, various keyword search semantics have been designed. However, these keyword search semantics and algorithms encounter efficiency or scalability issues. In this thesis, we propose new three generic frameworks or index techniques to address these issues. The thesis results show that the keyword search on massive graphs under different scenarios can be effective and efficient, which would facilitate keyword search services on graphs in the real world. First, we study the keyword search on massive knowledge graphs. In particular, we propose a generic ontology- based indexing framework for keyword search, called Bisimulation of Generalized Graph Index (BiG-index), to enhance the search performance. The novelties of BiG-index reside in using an ontology graph GOnt to summarize and index a data graph G iteratively, to form a hierarchical index structure G. Regarding query evaluation, we transform a keyword search q into Q according to GOnt in runtime. The transformed query is searched on the summary graphs in G. The efficiency is due to the small sizes of the summary graphs and the early pruning of semantically irrelevant subgraphs. To illustrate BiG-index's applicability, we show popular indexing techniques for keyword search can be easily implemented on top of BiG-index. Our extensive experiments show that BiG-index clearly reduced the runtimes of popular keyword search algorithms. Second, we study the problem of keyword search on public-private graph. In many applications (e.g., social networks), users may prefer to hide parts or all of her/his data graphs (e.g., private friendships) from the public. This leads to a recent graph model, namely the public-private network model, in which each user has his/her own network. While there have been studies on public-private network analysis, keyword search on public-private networks has not yet been studied. Hence, we propose a new keyword search framework, called public-private keyword search (PPKWS). PPKWS consists of three major steps: partial evaluation, answer refinement, and answer completion. We select three representative ones and show that they can be implemented on the model with minor modifications. We propose indexes and optimizations for PPKWS. We have verified through experiments that, on average, the algorithms implemented on top of PPKWS run 113 times faster than the original algorithms directly running on the public network attached to the private network for retrieving answers that span through them. Third, we study the keyword search in distributed graph evaluation systems. In the recent research on query evaluation, parallel evaluation has attracted much interest. However, the study on keyword search on distributed graphs has still been limited. We propose a novel distributed keyword search framework called DKWS. We propose a notify-push paradigm which can exchange the upper bounds of answers across all the workers asynchronously. In particular, the workers notify the coordinator when the local upper bound is refined. The coordinator pushes the refined global upper bound to all the workers. Moreover, we propose an efficient and generic keyword search algorithm for the workers. We have implemented DKWS on top of GRAPE, a distributed graph process system from our previous research collaboration. Extensive experimental results show that DKWS outperforms current-state-of-art techniques

In this dissertation, three research directions were explored to alleviate two major issues, i.e., the use of incorrect models and training/test condition mismatches, in the modeling frameworks of modern spoken keyword search (KWS) systems. Each of the three research directions, which include (i) data-efficient training processes, (ii) system optimization objectives, and (iii) data augmentation, utilizes different types and amounts of training resources in different ways to ameliorate the two issues of acoustic and language modeling in modern KWS systems. To be more specific, resource-dependent keyword modeling, keyword-boosted sMBR (state-level minimum Bayes risk) training, and multilingual acoustic modeling are proposed and investigated for acoustic modeling in this research. For language modeling, keyword-aware language modeling, discriminative keyword-aware language modeling, and web text augmented language modeling are presented and discussed. The dissertation provides a comprehensive collection of solutions and strategies to the acoustic and language modeling problems in KWS. It also offers insights into the realization of good-performance KWS systems. Experimental results show that the data-efficient training process and data augmentation are the two directions providing the most prominent performance improvement for KWS systems. While modifying system optimization objectives provides smaller yet consistent performance enhancement in KWS systems with different configurations. The effects of the proposed acoustic and language modeling approaches in the three directions are also shown to be additive and can be combined to further improve the overall KWS system performance.