A presente pesquisa analisou a recomendação de objetos de aprendizagem (OAs) por competências, a partir da colaboração baseada na avaliação pela qualidade percebida dos objetos de aprendizagem utilizados, considerando os estados de ânimo dos alunos. Neste sentido, a pesquisa visou desenvolver e validar um sistema de recomendação por qualidade percebida de objetos de aprendizagem por competência, que considera o estado de ânimo do aluno, denominado Requali. Este agregou os estados de ânimo dos usuários e a avaliação de qualidade percebida ao processo de recomendação de objetos de aprendizagem por competências. Os sistemas de recomendação integram processos que permitem caracterizar o perfil do usuário, as características do objeto e a avaliação do que está sendo disponibilizado. Assim, a fundamentação teórica inclui a recomendação de objetos de aprendizagem; os estados de ânimo dos usuários; e a avaliação de qualidade percebida. Para a recomendação de OAs, adotaram-se os estudos de Cazella, Nunes e Reategui (2010) sobre a relevância e as características dos sistemas de recomendação; e de Cazella et al. (2012) que tratam do RecOAComp como sistema de recomendação de objetos de aprendizagem por competências. Acerca dos estados de ânimo, incluíram-se o trabalho de Bercht (2001), sobre a essencialidade da afetividade no processo de ensino; e o de Longhi (2011), sobre o mapa afetivo como funcionalidade do Ambiente Virtual de Aprendizagem ROODA (Rede Cooperativa de Aprendizagem) para reconhecimento dos estados de ânimo. Por fim, utilizou-se a teoria da qualidade percebida desenvolvida por Oliver (1997), que trata das expectativas, das percepções e da avaliação da qualidade percebida pelos sujeitos. A pesquisa possibilitou a construção da estrutura de uma Matriz que foi utilizada posteriormente no Requali. Os critérios utilizados (expectativa, percepção e qualidade percebida) foram identificados na literatura e discutidos por um grupo focal de professores de Administração. O sistema Requali foi desenvolvido com tecnologia PHP, CSS e banco de dados MySQL. Ele foi validado por meio de um curso de extensão oferecido para alunos de graduação de Administração. Os alunos apontaram que o sistema fornece disponibilizações adequadas. Entretanto destacaram limitações de uso referentes ao reconhecimento de conceitos presentes na interface do sistema e na recuperação de informações, apesar de terem caracterizado o Requali como atrativo, fácil de usar e motivador. Assim, constatou-se que o Requali ao associar a avaliação pela qualidade percebida, baseada na Matriz Requali, e os estados de ânimos dos alunos para a recomendação de OAs por competências oferece itens disponibilizados que se aproximam dos interesses dos alunos. / This research analyzed the recommendation of learning objects (LO) by competence from the contribution based on the perceived quality evaluation of the learning objects that have been used, considering students’ state of mood. In this sense, the research aimed at developing and validating a recommendation system by perceived quality of learning objects, which considers students’ state of mood, that is named Requali. This system has associated users’ states of mood and perceived evaluation in the recommendation process of learning objects by competencies. Recommender systems integrate processes that allow the characterization of the user’s profile, the object’s characteristics and the evaluation of what is made available. Thus the theoretical framework includes learning objects recommendation; users’ states of mood; and the assessment of the perceived quality. For LO recommendation, we adopted the studies of Cazella, Nunes and Reategui (2010) on the relevance and characteristics of recommender systems; and the studies of Cazella et al. (2012) that approaches the RecOAComp as a learning objects recommender system by competencies. Regarding states of mood, we included the work of Bercht (2001) on the essentiality of affectivity in the teaching process; and Longhi’s work (2011) on the emotional map as a functionality of the ROODA (Learning Cooperative Network) Learning Virtual Environment for states of mood recognition. Finally, we used the perceived quality theory developed by Oliver (1997), which approaches the subject’s expectations, perceptions and assessment of the perceived quality. The research enabled the construction of the structure of a matrix which was subsequently used in Requali. The used criteria (expectation, perception and perceived quality) were identified in the literature and discussed by a focus group with business administration professors. Requali system was developed with PHP, CSS and MySQL database. It has been validated by an extension course offered for Business Administration undergraduate students. The students have pointed out that the system provides appropriate recommendations. However, they have highlighted limitations of use regarding concepts recognition in the system’s interface and information retrieval, although they have also characterized Requali as attractive, easy to use and motivating. Thus it was found that Requali, associating perceived quality evaluation based on the Requali Matrix with the students’ states of mood for LO recommendation by competencies,offers available items that are close to student’s interests. / Esta investigación examinó las recomendaciones de los objetos de aprendizaje de competencias (los) de la colaboración basada en la evaluación de la calidad percibida de los objetos utilizados aprendizaje, teniendo en cuenta los estados de ánimo de los estudiantes. En este sentido, la investigación tuvo como objetivo desarrollar y validar un sistema de recomendación de calidad para la competencia percibida por los objetos de aprendizaje, de los objetos de aprendizaje (OAs) por competencia, que considera el estado de ánimo del estudiante llamado Requali. Este añadió los estados de ánimo de los usuarios y la evaluación de la calidad percibida en el proceso de recomendación de los objetos de aprendizaje por competencias. Los sistemas de recomendación se integran procesos que permiten caracterizar el perfil de lo usuario, las características del objeto y la evaluación de lo que se está poniendo a disposición. Así, el marco teórico incluye la recomendación de objetos de aprendizaje; estados de ánimo de los usuários; y la evaluación de la calidad percibida. Hacia la recomendación de los OAs, fueron adoptdos los estudios de Cazella, Nunes y Reátegui (2010) acerca de la importancia y las características de los sistemas de recomendación; y Cazella et al. (2012), relativo a ló RecOAComp como sistema de recomendación de objetos de aprendizaje por competencias. En cuanto a los estados de ánimo, se incluye el trabajo de Bercht (2001) sobre la esencialidad de la afectividad en el proceso de enseñanza; y Longhi (2011), sobre el mapa afectivo como la funcionalidad del Entorno Virtual de Aprendizaje ROODA (Red de Aprendizaje Cooperativa) para el reconocimiento de los estados de ánimo. Por último, se utilizó la teoría de la calidad percibida desarrollada por Oliver (1997), que se ocupa de las expectativas, las percepciones y la evaluación de la calidad percibida por los sujetos. La investigación permitió la construcción de la estructura de una matriz que fuera utilizada posteriormente en ló Requali. Los critérios utilizados (expectativa, la percepción y la calidad percibida) fueron identificados en la literatura y discutidos por um grupo focal de profesores de Gestión. El sistema Requali fue desarrollado con tecnología PHP, CSS y banco de datos MySQL. Fue validado por un curso de extensión ofrecido para estudiantes universitarios de Administración. Los estudiantes señalaron que el sistema proporciona disponibilizaciones apropiadas. Sin embargo han resaltado las limitaciones de uso para el reconocimiento de los conceptos presentes en la interfaz del sistema y la recuperación de información a pesar de que caracterizarán lo Requali como atractivo, fácil de usar y motivador. Por lo tanto, se encontró que el Requali asociando la evaluación por parte de la calidad percibida en base a la matriz Requali, y los estados de ánimo de los estudiantes para la recomendación de OAs por la competencia ofrece los elementos disponibles que se cercan de los intereses de los estudiantes.

Au cours des dernières années, le Web a connu une énorme croissance en matière de contenus et d'utilisateurs. Ce phénomène a entraîné des problèmes liés à la surcharge d'information face à laquelle les utilisateurs ont des difficultés à trouver les bonnes informations. Des systèmes de recommandation ont été développés pour résoudre ce problème afin de guider les utilisateurs dans ce flux d'informations. Les approches de recommandation se sont multipliées et ont été mises en œuvre avec succès, notamment au travers d’approches telles que le filtrage collaboratif. Cependant, il existe encore des défis et des limites qui offrent des opportunités pour de nouvelles recherches. Parmi ces défis, la conception de systèmes de recommandation de lectures est devenue un axe de recherche en pleine expansion suite à l’apparition des bibliothèques numériques.Traditionnellement, les bibliothèques jouent un rôle passif dans l’interaction avec les lecteurs et ce, faute d’outils efficaces de recherche et de recommandation. Dans ce manuscrit, nous nous sommes penchée sur la création d’un système de recommandation de lectures. Nos objectifs portent sur :- améliorer la compréhension des besoins utilisateurs exprimés au sein des requêtes en langage naturel de recherches de livres, articles et billets ; - pallier l'absence de liens explicites entre ouvrages et articles de revues par la détection et l'analyse automatique des références bibliographiques afin de proposer des liens ; - parvenir à un système de recommandation de lectures s'appuyant sur des données textuelles permettant de fournir une liste de recommandations personnalisées aux utilisateurs actifs. / In the recent years, the Web has undergone a tremendous growth regarding both content and users. This has led to an information overload problem in which people are finding it increasingly difficult to locate the right information at the right time. Recommender systems have been developed to address this problem, by guiding users through the big ocean of information. The recommendation approaches have multiplied and have been successfully implemented, particularly through approaches such as collaborative filtering. However, there are still challenges and limitations that offer opportunities for new research. Among these challenges, the design of reading recommendation systems has become a new expanding research focus following the emergence of digital libraries.Traditionally, libraries play a passive role in interaction with users due to the lack of effective search and recommendation tools. In this manuscript, we will study the creation of a reading recommendation system in which we'll try to exploit the possibilities of digital access to scientific information. Our objectives are: - to improve the understanding of user needs expressed in natural language search queries for books, articles and posts. This work will require the establishment of processes capable of exploiting the structures of data and their dimension; - to compensate for the absence of explicit links between books and journal articles by automatically detecting and analyzing bibliographic references, and then to propose links;- to achieve a reading recommendation system based on textual data to provide a customized recommendation list to active users, similar to systems already used by users profiles.

Les technologies de l'information et le succès des services associés (forums, sites spécialisés, etc) ont ouvert la voie à un mode d'expression massive d'opinions sur les sujets les plus variés (e-commerce, critiques artistiques, etc). Cette profusion d'opinions constitue un véritable eldorado pour l'internaute, mais peut rapidement le conduire à une situation d'indécision car les avis déposés peuvent être fortement disparates voire contradictoires. Pour une gestion fiable et pertinente de l'information contenue dans ces avis, il est nécessaire de mettre en place des systèmes capables de traiter directement les opinions exprimées en langage naturel afin d'en contrôler la subjectivité et de gommer les effets de lissage des traitements statistiques. La plupart des systèmes dits de recommandation ne prennent pas en compte toute la richesse sémantique des critiques et leur associent souvent des systèmes d'évaluation qui nécessitent une implication conséquente et des compétences particulières chez l'internaute. Notre objectif est de minimiser l'intervention humaine dans le fonctionnement collaboratif des systèmes de recommandation en automatisant l'exploitation des données brutes que constituent les avis en langage naturel. Notre approche non supervisée de segmentation thématique extrait les sujets d'intérêt des critiques, puis notre technique d'analyse de sentiments calcule l'opinion exprimée sur ces critères. Ces méthodes d'extraction de connaissances combinées à des outils d'analyse multicritère adaptés à la fusion d'avis d'experts ouvrent la voie à des systèmes de recommandation pertinents, fiables et personnalisés. / Information Technology and the success of its related services (blogs, forums, etc.) have paved the way for a massive mode of opinion expression on the most varied subjects (e-commerce websites, art reviews, etc). This abundance of opinions could appear as a real gold mine for internet users, but it can also be a source of indecision because available opinions may be ill-assorted if not contradictory. A reliable and relevant information management of opinions bases requires systems able to directly analyze the content of opinions expressed in natural language. It allows controlling subjectivity in evaluation process and avoiding smoothing effects of statistical treatments. Most of the so-called recommender systems are unable to manage all the semantic richness of a review and prefer to associate to the review an assessment system that supposes a substantial implication and specific competences of the internet user. Our aim is minimizing user intervention in the collaborative functioning of recommender systems thanks to an automated processing of available reviews in natural language by the recommender system itself. Our topic segmentation method extracts the subjects of interest from the reviews, and then our sentiment analysis approach computes the opinion related to these criteria. These knowledge extraction methods are combined with multicriteria analysis techniques adapted to expert assessments fusion. This proposal should finally contribute to the coming of a new generation of more relevant, reliable and personalized recommender systems.

No description available.

A presente pesquisa analisou a recomendação de objetos de aprendizagem (OAs) por competências, a partir da colaboração baseada na avaliação pela qualidade percebida dos objetos de aprendizagem utilizados, considerando os estados de ânimo dos alunos. Neste sentido, a pesquisa visou desenvolver e validar um sistema de recomendação por qualidade percebida de objetos de aprendizagem por competência, que considera o estado de ânimo do aluno, denominado Requali. Este agregou os estados de ânimo dos usuários e a avaliação de qualidade percebida ao processo de recomendação de objetos de aprendizagem por competências. Os sistemas de recomendação integram processos que permitem caracterizar o perfil do usuário, as características do objeto e a avaliação do que está sendo disponibilizado. Assim, a fundamentação teórica inclui a recomendação de objetos de aprendizagem; os estados de ânimo dos usuários; e a avaliação de qualidade percebida. Para a recomendação de OAs, adotaram-se os estudos de Cazella, Nunes e Reategui (2010) sobre a relevância e as características dos sistemas de recomendação; e de Cazella et al. (2012) que tratam do RecOAComp como sistema de recomendação de objetos de aprendizagem por competências. Acerca dos estados de ânimo, incluíram-se o trabalho de Bercht (2001), sobre a essencialidade da afetividade no processo de ensino; e o de Longhi (2011), sobre o mapa afetivo como funcionalidade do Ambiente Virtual de Aprendizagem ROODA (Rede Cooperativa de Aprendizagem) para reconhecimento dos estados de ânimo. Por fim, utilizou-se a teoria da qualidade percebida desenvolvida por Oliver (1997), que trata das expectativas, das percepções e da avaliação da qualidade percebida pelos sujeitos. A pesquisa possibilitou a construção da estrutura de uma Matriz que foi utilizada posteriormente no Requali. Os critérios utilizados (expectativa, percepção e qualidade percebida) foram identificados na literatura e discutidos por um grupo focal de professores de Administração. O sistema Requali foi desenvolvido com tecnologia PHP, CSS e banco de dados MySQL. Ele foi validado por meio de um curso de extensão oferecido para alunos de graduação de Administração. Os alunos apontaram que o sistema fornece disponibilizações adequadas. Entretanto destacaram limitações de uso referentes ao reconhecimento de conceitos presentes na interface do sistema e na recuperação de informações, apesar de terem caracterizado o Requali como atrativo, fácil de usar e motivador. Assim, constatou-se que o Requali ao associar a avaliação pela qualidade percebida, baseada na Matriz Requali, e os estados de ânimos dos alunos para a recomendação de OAs por competências oferece itens disponibilizados que se aproximam dos interesses dos alunos. / This research analyzed the recommendation of learning objects (LO) by competence from the contribution based on the perceived quality evaluation of the learning objects that have been used, considering students’ state of mood. In this sense, the research aimed at developing and validating a recommendation system by perceived quality of learning objects, which considers students’ state of mood, that is named Requali. This system has associated users’ states of mood and perceived evaluation in the recommendation process of learning objects by competencies. Recommender systems integrate processes that allow the characterization of the user’s profile, the object’s characteristics and the evaluation of what is made available. Thus the theoretical framework includes learning objects recommendation; users’ states of mood; and the assessment of the perceived quality. For LO recommendation, we adopted the studies of Cazella, Nunes and Reategui (2010) on the relevance and characteristics of recommender systems; and the studies of Cazella et al. (2012) that approaches the RecOAComp as a learning objects recommender system by competencies. Regarding states of mood, we included the work of Bercht (2001) on the essentiality of affectivity in the teaching process; and Longhi’s work (2011) on the emotional map as a functionality of the ROODA (Learning Cooperative Network) Learning Virtual Environment for states of mood recognition. Finally, we used the perceived quality theory developed by Oliver (1997), which approaches the subject’s expectations, perceptions and assessment of the perceived quality. The research enabled the construction of the structure of a matrix which was subsequently used in Requali. The used criteria (expectation, perception and perceived quality) were identified in the literature and discussed by a focus group with business administration professors. Requali system was developed with PHP, CSS and MySQL database. It has been validated by an extension course offered for Business Administration undergraduate students. The students have pointed out that the system provides appropriate recommendations. However, they have highlighted limitations of use regarding concepts recognition in the system’s interface and information retrieval, although they have also characterized Requali as attractive, easy to use and motivating. Thus it was found that Requali, associating perceived quality evaluation based on the Requali Matrix with the students’ states of mood for LO recommendation by competencies,offers available items that are close to student’s interests. / Esta investigación examinó las recomendaciones de los objetos de aprendizaje de competencias (los) de la colaboración basada en la evaluación de la calidad percibida de los objetos utilizados aprendizaje, teniendo en cuenta los estados de ánimo de los estudiantes. En este sentido, la investigación tuvo como objetivo desarrollar y validar un sistema de recomendación de calidad para la competencia percibida por los objetos de aprendizaje, de los objetos de aprendizaje (OAs) por competencia, que considera el estado de ánimo del estudiante llamado Requali. Este añadió los estados de ánimo de los usuarios y la evaluación de la calidad percibida en el proceso de recomendación de los objetos de aprendizaje por competencias. Los sistemas de recomendación se integran procesos que permiten caracterizar el perfil de lo usuario, las características del objeto y la evaluación de lo que se está poniendo a disposición. Así, el marco teórico incluye la recomendación de objetos de aprendizaje; estados de ánimo de los usuários; y la evaluación de la calidad percibida. Hacia la recomendación de los OAs, fueron adoptdos los estudios de Cazella, Nunes y Reátegui (2010) acerca de la importancia y las características de los sistemas de recomendación; y Cazella et al. (2012), relativo a ló RecOAComp como sistema de recomendación de objetos de aprendizaje por competencias. En cuanto a los estados de ánimo, se incluye el trabajo de Bercht (2001) sobre la esencialidad de la afectividad en el proceso de enseñanza; y Longhi (2011), sobre el mapa afectivo como la funcionalidad del Entorno Virtual de Aprendizaje ROODA (Red de Aprendizaje Cooperativa) para el reconocimiento de los estados de ánimo. Por último, se utilizó la teoría de la calidad percibida desarrollada por Oliver (1997), que se ocupa de las expectativas, las percepciones y la evaluación de la calidad percibida por los sujetos. La investigación permitió la construcción de la estructura de una matriz que fuera utilizada posteriormente en ló Requali. Los critérios utilizados (expectativa, la percepción y la calidad percibida) fueron identificados en la literatura y discutidos por um grupo focal de profesores de Gestión. El sistema Requali fue desarrollado con tecnología PHP, CSS y banco de datos MySQL. Fue validado por un curso de extensión ofrecido para estudiantes universitarios de Administración. Los estudiantes señalaron que el sistema proporciona disponibilizaciones apropiadas. Sin embargo han resaltado las limitaciones de uso para el reconocimiento de los conceptos presentes en la interfaz del sistema y la recuperación de información a pesar de que caracterizarán lo Requali como atractivo, fácil de usar y motivador. Por lo tanto, se encontró que el Requali asociando la evaluación por parte de la calidad percibida en base a la matriz Requali, y los estados de ánimo de los estudiantes para la recomendación de OAs por la competencia ofrece los elementos disponibles que se cercan de los intereses de los estudiantes.

推薦系統的出現是為了解決訊息過載的問題，其需求隨著科技的進步、網路的普及而增加，相關技術也越發多樣且成熟。廣泛應用於各領域的統計模型也在技術的行列中。 推薦系統的運作仰賴使用者偏好訊息，而使用者對項目所組成的偏好空間往往十分巨大且不平衡，統計上需要相對複雜的隨機效應模型或混合效應模型來描繪這樣的變數結構，且通常需要計算效率相對低的反覆疊代過程來估計模型參數。因此Perry（2014）、Gao & Owen（2016）先後提出以動差法處理階層線性模型與兩因子隨機效應模型，是一種犧牲統計效率換取計算效率的做法。 本研究便是採用統計模型中的隨機效應模型法，分別以最大概似法和動差法估計參數，與同為協同過濾技術觀點的矩陣分解法進行分析比較。透過預測準確度和運算效率兩個層面，來評估各演算法在MoiveLens這筆資料上的推薦表現。 根據試驗結果歸納出隨機效應模型法無論以什麼樣的參數估計方式，在預測準確度的表現上都不如矩陣分解法來得好；但以動差法估計參數在穩定度上與矩陣分解法的表現差不多，且在運算效率上好很多。 / The recommender system (RS) appeared to solve the problem of information overload. The demand of the RS has increased with the advancement of technology and the popularity of the Internet, and related techniques have become more diverse and mature. The statistical models widely used in various fields are also in the list of techniques. The operation of the RS relies on user preference information, and the space of users’ preference to items is often large and unbalanced. Statistically, relatively complex random effects models or mixed effects models are needed to describe such variable structures, and often require a large number of iterations to estimate model parameters. Perry (2014), Gao & Owen (2016) proposed using the moment-based method to deal with hierarchical linear models and two-factor random effects models, respectively, expressing an idea of sacrificing statistical efficiency in exchange for computational efficiency. In this study, we analyze and compare the random effects model, using the maximum likelihood method and the moment-based method to estimate the parameters with the matrix factorization. Through the prediction accuracy and computational efficiency to evaluate the performance of each algorithm on the MoiveLens data. According to the experiment results, the random effects model is not as good as the matrix factorization in terms of the prediction accuracy no matter what kind of parameter estimation method is used; however, the performance of the moment-based parameter estimation is consistent with the matrix factorization in terms of the prediction stability, and much better in terms of the efficiency.

The aim of this work is to identify the possibilities of the complementary usage of two analytical methods of data analysis, OLAP analysis and data mining represented by GUHA association rule mining. The usage of these two methods in the context of proposed scenarios on one dataset presumes a synergistic effect, surpassing the knowledge acquired by these two methods independently. This is the main contribution of the work. Another contribution is the original use of GUHA association rules where the mining is performed on aggregated data. In their abilities, GUHA association rules outperform classic association rules referred to the literature. The experiments on real data demonstrate the finding of unusual trends in data that would be very difficult to acquire using standard methods of OLAP analysis, the time consuming manual browsing of an OLAP cube. On the other hand, the actual use of association rules loses a general overview of data. It is possible to declare that these two methods complement each other very well. The part of the solution is also usage of LMCL scripting language that automates selected parts of the data mining process. The proposed recommender system would shield the user from association rules, thereby enabling common analysts ignorant of the association rules to use their possibilities. The thesis combines quantitative and qualitative research. Quantitative research is represented by experiments on a real dataset, proposal of a recommender system and implementation of the selected parts of the association rules mining process by LISp-Miner Control Language. Qualitative research is represented by structured interviews with selected experts from the fields of data mining and business intelligence who confirm the meaningfulness of the proposed methods.

Notre recherche porte sur la recommandation de nouvelles offres d'emploi venant d'être postées et n'ayant pas d'historique d'interactions (démarrage à froid). Nous adaptons les systèmes de recommandations bien connus dans le domaine du commerce électronique à cet objectif, en exploitant les traces d'usage de l'ensemble des demandeurs d'emploi sur les offres antérieures. Une des spécificités du travail présenté est d'avoir considéré des données réelles, et de s'être attaqué aux défis de l'hétérogénéité et du bruit des documents textuels. La contribution présentée intègre l'information des données collaboratives pour apprendre une nouvelle représentation des documents textes, requise pour effectuer la recommandation dite à froid d'une offre nouvelle. Cette représentation dite latente vise essentiellement à construire une bonne métrique. L'espace de recherche considéré est celui des réseaux neuronaux. Les réseaux neuronaux sont entraînés en définissant deux fonctions de perte. La première cherche à préserver la structure locale des informations collaboratives, en s'inspirant des approches de réduction de dimension non linéaires. La seconde s'inspire des réseaux siamois pour reproduire les similarités issues de la matrice collaborative. Le passage à l'échelle de l'approche et ses performances reposent sur l'échantillonnage des paires d'offres considérées comme similaires. L'intérêt de l'approche proposée est démontrée empiriquement sur les données réelles et propriétaires ainsi que sur le benchmark publique CiteULike. Enfin, l'intérêt de la démarche suivie est attesté par notre participation dans un bon rang au challenge international RecSys 2017 (15/100; un million d'utilisateurs pour un million d'offres). / Our research focuses on the recommendation of new job offers that have just been posted and have no interaction history (cold start). To this objective, we adapt well-knowns recommendations systems in the field of e-commerce by exploiting the record of use of all job seekers on previous offers. One of the specificities of the work presented is to have considered real data, and to have tackled the challenges of heterogeneity and noise of textual documents. The presented contribution integrates the information of the collaborative data to learn a new representation of text documents, which is required to make the so-called cold start recommendation of a new offer. The new representation essentially aims to build a good metric. The search space considered is that of neural networks. Neural networks are trained by defining two loss functions. The first seeks to preserve the local structure of collaborative information, drawing on non-linear dimension reduction approaches. The second is inspired by Siamese networks to reproduce the similarities from the collaborative matrix. The scaling up of the approach and its performance are based on the sampling of pairs of offers considered similar. The interest of the proposed approach is demonstrated empirically on the real and proprietary data as well as on the CiteULike public benchmark. Finally, the interest of the approach followed is attested by our participation in a good rank in the international challenge RecSys 2017 (15/100, with millions of users and millions of offers).

Mit der steigenden Verfügbarkeit komponenten- und serviceorientiert bereitgestellter Ressourcen und Dienstleistungen entwickelt sich das Web zu einer geeigneten Plattform für vielfältige Anwendungsszenarien. Darauf aufbauend entstehen komposite Webanwendungen durch das Rekombinieren und Verknüpfen vorhandener Bausteine. Auf diese Weise kann ein funktionaler Mehrwert zur Lösung situationsspezifischer Problemstellungen erzielt werden. Zunehmend wird angestrebt, dass Endnutzer selbst als Anwendungsentwickler in Erscheinung treten. Dieses Prinzip, das End-User-Development, ist ökonomisch lukrativ, da Nischenanforderungen effizienter erfüllt werden können. Allerdings stehen dabei insbesondere Domänenexperten ohne Programmiererkenntnisse noch immer vor substantiellen Herausforderungen, wie der bedarfsgerechten Auswahl von Bausteinen und deren korrekten Komposition. Diese Dissertation stellt daher neue Methoden und Werkzeuge für das assistierte End-User-Development von kompositen Webanwendungen vor. Im Ergebnis entsteht das ganzheitliche Konzept einer Kompositionsplattform, die Nicht-Programmierer in die Lage versetzt, eigenständig Anwendungen bedarfsgerecht zu entwickeln und einzusetzen. Als zentrales Element existiert ein hochiteratives Vorgehensmodell, bei dem die Entwicklung und die Nutzung kompositer Webanwendungen weitgehend verschmelzen. Ein wesentliches Merkmal des Ansatzes ist, dass aus Nutzersicht sämtliche Aktivitäten auf fachlicher Ebene stattfinden, während die Kompositionsplattform die technische Umsetzung übernimmt und vor den Nutzern verbirgt. Grundlage hierfür sind Konzepte zur universellen Komposition und eine umfassende Modellbasis. Letztere umfasst semantikbasierte Beschreibungen von Komponenten sowie Kompositionsfragmenten und von deren Funktionalitäten (Capabilities). Weiterhin wird statistisches und semantisches Kompositionswissen sowie Nutzerfeedback modelliert. Darauf aufbauend werden neue, anwendungsunabhängige Mechanismen konzipiert. Hierzu zählt ein Empfehlungssystem, das prozessbegleitend Kompositionsschritte vorschlägt und das erstmals mit Empfehlungsstrategien in hohem Maße an seinen Einsatzkontext angepasst werden kann. Weiterhin sieht der Ansatz semantikbasierte Datenmediation und einen Algorithmus vor, der die Capabilities von Kompositionsfragmenten abschätzt. Diese Konzepte dienen schließlich als Basis für eine in sich zusammenhängende Werkzeugpalette, welche die Aktivitäten des Vorgehensmodells durchgehend unterstützt. Zum Beispiel assistiert ein Wizard Nicht-Programmierern bei der anforderungsgetriebenen Identifikation passender Kompositionsfragmente. Weitere konzipierte Hilfsmittel erlauben es Nutzern, Anwendungen live zu komponieren sowie anzupassen und deren Funktionsweise nachzuvollziehen bzw. zu untersuchen. Die Werkzeuge basieren maßgeblich auf Capabilities zur fachlichen Kommunikation mit Nutzern, als Kompositionsmetapher, zur Erklärung funktionaler Zusammenhänge und zur Erfassung von Nutzeranforderungen. Die Kernkonzepte wurden durch prototypische Implementierungen und praktische Erprobung in verschiedenen Anwendungsdomänen validiert. Zudem findet die Evaluation von Ansätzen durch Performanz-Messungen, Expertenbefragung und Nutzerstudien statt. Insgesamt zeigen die Ergebnisse, dass die Konzepte für die Zielgruppe nützlich sind und als tragfähig angesehen werden können.:1 Einleitung 1.1 Analyse von Herausforderungen und Problemen 1.1.1 Zielgruppendefinition 1.1.2 Problemanalyse 1.2 Thesen, Ziele, Abgrenzung 1.2.1 Forschungsthesen 1.2.2 Forschungsziele 1.2.3 Annahmen und Abgrenzungen 1.3 Aufbau der Arbeit 2 Grundlagen und Anforderungsanalyse 2.1 CRUISE – Architektur und Modelle 2.1.1 Komponentenmetamodell 2.1.2 Kompositionsmodell 2.1.3 Architekturüberblick 2.1.4 Fazit 2.2 Referenzszenarien 2.2.1 Ad-hoc-Erstellung einer CWA zur Konferenzplanung 2.2.2 Geführte Recherche nach einer CWA 2.2.3 Unterstützte Nutzung einer CWA 2.3 Anforderungen 3 Stand von Forschung und Technik 3.1 Kompositionsplattformen für EUD 3.1.1 Webservice-Komposition durch Endnutzer 3.1.2 Mashup-Komposition durch Endnutzer 3.1.3 Fazit 3.2 Empfehlungssysteme im Mashupkontext 3.2.1 Empfehlungsansätze in Kompositionsplattformen 3.2.2 Nutzerfeedback in Empfehlungssystemen 3.2.3 Fazit 3.3 Eingabe funktionaler Anforderungen 3.3.1 Textuelle Ansätze 3.3.2 Graphische Anfrageformulierung 3.3.3 Hierarchische und facettierte Suche 3.3.4 Assistenten und dialogbasierte Ansätze 3.3.5 Fazit 3.4 Ansätze zur Datenmediation 3.4.1 Ontology Mediation 3.4.2 Vertreter aus dem Bereich (Semantic) Web Services 3.4.3 Datenmediation in Mashup-Plattformen 3.4.4 Fazit 3.5 Fazit zum Stand von Forschung und Technik 4 Assistiertes EUD von CWA durch Nicht-Programmierer 4.1 Assistiertes EUD von Mashups 4.1.1 Modellebene 4.1.2 Basismechanismen 4.1.3 Werkzeuge 4.2 Grobarchitektur 5 Basiskonzepte 5.1 Grundlegende Modelle 5.1.1 Capability-Metamodell 5.1.2 Erweiterungen von Komponentenmodell und SMCDL 5.1.3 Nutzer- und Kontextmodell 5.1.4 Metamodell für kontextualisiertes Feedback 5.2 Semantische Datenmediation 5.2.1 Vorbetrachtungen und Definitionen 5.2.2 Techniken zur semantischen Datenmediation 5.2.3 Architektonische Implikationen und Abläufe 5.3 Ableiten von Capabilities 5.3.1 Anforderungen und verwandte Ansätze 5.3.2 Definitionen und Grundlagen 5.3.3 Übersicht über den Algorithmus 5.3.4 Detaillierter Ablauf 5.3.5 Architekturüberblick 5.4 Erzeugung eines Capability-Wissensgraphen 5.4.1 Struktur des Wissensgraphen 5.4.2 Instanziierung des Wissensgraphen 5.5 Zusammenfassung 6 Empfehlungssystem 6.1 Gesamtansatz im Überblick 6.2 Empfehlungssystemspezifische Metamodelle 6.2.1 Trigger-Metamodell 6.2.2 Pattern-Metamodell 6.3 Architektur und Abläufe des Empfehlungssystems 6.3.1 Ableitung von Pattern-Instanzen 6.3.2 Empfehlungsgründe identifizieren durch Trigger 6.3.3 Empfehlungen berechnen 6.3.4 Präsentation von Empfehlungen 6.3.5 Integration von Patterns 6.4 Zusammenfassung 7 Methoden zur Nutzerführung 7.1 Der Startbildschirm als zentraler Einstiegspunkt 7.2 Live-View 7.3 Capability-View 7.3.1 Interaktive Exploration von Capabilities 7.3.2 Kontextsensitive Erzeugung von Beschriftungen 7.3.3 Verknüpfen von Capabilities 7.3.4 Handhabung von Komponenten ohne UI 7.4 Wizard zur Eingabe funktionaler Anforderungen 7.5 Erklärungstechniken 7.5.1 Anforderungen und verwandte Ansätze 7.5.2 Kernkonzepte 7.5.3 Assistenzwerkzeuge 8 Implementierung und Evaluation 8.1 Umsetzung der Modelle und der Basisarchitektur 8.2 Realisierung der Mediationskonzepte 8.2.1 Erweiterung des Kompositionsmodells 8.2.2 Implementierung des Mediators 8.2.3 Evaluation und Diskussion 8.3 Algorithmus zur Abschätzung von Capabilities 8.3.1 Prototypische Umsetzung 8.3.2 Experten-Evaluation 8.4 Umsetzung des Empfehlungskreislaufes 8.4.1 Performanzbetrachtungen 8.4.2 Evaluation und Diskussion 8.5 Evaluation von EUD-Werkzeugen 8.5.1 Evaluation der Capability-View 8.5.2 Prototyp und Nutzerstudie des Wizards 8.5.3 Prototyp und Nutzerstudie zu den Erklärungstechniken 8.6 Fazit 9 Zusammenfassung, Diskussion und Ausblick 9.1 Zusammenfassung und Beiträge der Kapitel 9.2 Einschätzung der Ergebnisse 9.2.1 Diskussion der Erreichung der Forschungsziele 9.2.2 Diskussion der Forschungsthesen 9.2.3 Wissenschaftliche Beiträge 9.2.4 Grenzen der geschaffenen Konzepte 9.3 Laufende und weiterführende Arbeiten A Anhänge A.1 Richtlinien für die Annotation von Komponenten A.2 Fragebogen zur System Usability Scale A.3 Illustration von Mediationstechniken A.4 Komponentenbeschreibung in SMCDL (Beispiel) A.5 Beispiele zu Algorithmen A.5.1 Berechnung einer bestimmenden Entity A.5.2 Berechnung der Ähnlichkeit atomarer Capabilities A.6 Bewertung verwandter Ansätze Literaturverzeichnis Webreferenzen

In our daily life, time is of the essence. People do not have time to browse through hundreds of thousands of digital items every day to find the right item for them. This is where a recommendation system shines. Tigerhall is a company that distributes podcasts, ebooks and events to subscribers. They are expanding their digital content warehouse which leads to more data for the users to filter. To make it easier for users to find the right podcast or the most exciting e-book or event, a recommendation system has been implemented. A recommender system can be implemented in many different ways. There are content-based filtering methods that can be used that focus on information about the items and try to find relevant items based on that. Another alternative is to use collaboration filtering methods that use information about what the consumer has previously consumed in correlation with what other users have consumed to find relevant items. In this project, a hybrid recommender system that uses a k-nearest neighbors algorithm alongside a matrix factorization algorithm has been implemented. The k-nearest neighbors algorithm performed well despite the sparse data while the matrix factorization algorithm performs worse. The matrix factorization algorithm performed well when the user has consumed plenty of items.