Logo des Repositoriums

ABIS 2010 – 18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond

Kassel, 4.-6. October 2010
Homepage des Workshops

Autor*innen mit den meisten Dokumenten  

Auflistung nach:

Neueste Veröffentlichungen

1 - 9 von 9
  • Konferenzbeitrag
    How Predictable Are You? A Comparison of Prediction Algorithms for Web Page Revisitation
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Kawase, Ricardo; Papadakis, George; Herder, Eelco
    Users return to Web pages for various reasons. Apart from pages visited due to backtracking, users typically monitor a number of favorite pages, while dealing with tasks that reoccur on an infrequent basis. In this paper, we introduce a novel method for predicting the next revisited page in a certain user context that, unlike existing methods, doesn’t rely on machine learning algorithms. We evaluate it over a large data set comprising the navigational activity of 25 users over a period of 6 months. The outcomes suggest a significant improvement over methods typically used in this context, thus paving the way for exploring new means of improving user’s navigational support.
  • Konferenzbeitrag
    User Models meet Digital Object Memories in the Internet of Things
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Heckmann, Dominik
    In this paper, we argue that Digital Object Memories in the Internet of Things are closely related to partial user models in Personalization and that research can gain insights by analogy from both sides. We describe UbisMemory, a Semantic Web middleware for partial user models and digital object memories. We describe the content representation, the service and its technical issues. We argue that Digital Object Memories can be extended and merged with “Digital User Memories” or life-long user models. We argue that Life-Logging for objects and for humans are closer related than expected.
  • Konferenzbeitrag
    On the Role of Social Tags in Filtering Interesting Resources from Folksonomies
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Godoy, Daniela
    Social tagging systems allow users to easily create, organize and share collections of resources (e.g. Web pages, research papers, photos, etc.) in a collaborative fashion. The rise in popularity of these systems in recent years go along with an rapid increase in the amount of data contained in their underlying folksonomies, thereby hindering the user task of discovering interesting resources. In this paper the problem of filtering resources from social tagging systems according to individual user interests using purely tagging data is studied. One-class classification is evaluated as a means to learn how to identify relevant information based on positive examples exclusively, since it is assumed that users expressed their interest in resources by annotating them while there is not an straightforward method to collect non-interesting information. The results of using social tags for personal classification are compared with those achieved with traditional information sources about the user interests such as the textual content of Web documents. Finding interesting resources based on social tags is an important benefit of exploiting the collective knowledge generated by tagging activities. Experimental evaluation showed that tag-based classification outperformed classifiers learned using the full-text of documents as well as other content-related sources.
  • Konferenzbeitrag
    What is wrong with the IMS Learning Design specification? Constraints And Recommendations
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Burgos, Daniel
    The work presented in this paper summarizes the research performed in order to implement a set of Units of Learning (UoLs) focused on adaptive learning processes, using the specification IMS Learning Design (IMS-LD). Through the implementation and analysis of four learning scenarios, and one additional application case, we identify a number of constraints on the use of IMS-LD to support adaptive learning. Indeed, our work in this paper shows how IMS-LD expresses adaptation. In addition, our research presents a number of elements and features that should be improved and-or modified to achieve a better support of adaptation for learning processes. Furthermore, we point out to interoperability and authoring issues too. Finally, we use the work carried out to suggest extensions and modifications of IMS-LD with the final aim of better supporting the implementation of adaptive learning processes.
  • Konferenzbeitrag
    Student Model Adjustment Through Random-Restart Hill Climbing
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Doost, Ahmad Salim; Melis, Erica
    ACTIVEMATH is a web-based intelligent tutoring system (ITS) for studying mathematics. Its course generator, which assembles content to personalized books, strongly depends on the underlying student model. Therefore, a student model is important to make an ITS adaptive. The more accurate it is, the better could be the adaptation. Here we present which parameters can be optimized and how they can be optimized in an efficient and affordable manner. This methodology can be generalized beyond ACTIVEMATH’s student model. We also present our results for the optimization based on two sets of log data. Our optimization method is based on random-restart hill climbing and it considerably improved the student model’s accuracy.
  • Konferenzbeitrag
    User and Document Group Approach of Clustering in Tagging Systems
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Pan, Rong; Xu, Guandong; Dolog, Peter
    In this paper, we propose a spectral clustering approach for users and documents group modeling in order to capture the common preference and relatedness of users and documents, and to reduce the time complexity of similarity calculations. In experiments, we investigate the selection of the optimal amount of clusters. We also show a reduction of the time consuming in calculating the similarity for the recommender systems by selecting a centroid first, and then compare the inside item on behalf of each group.
  • Konferenzbeitrag
    Modeling, obtaining and storing data from social media tools with Artefact-Actor-Networks
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Reinhardt, Wolfgang; Varlemann, Tobias; Moi, Matthias; Wilke, Adrian
    Social interaction between people has peerlessly changed with the availability of the Internet and the World Wide Web. The Internet brought new ways of communication technologies to live and enhanced people’s reachability, augmented possibilities for personal presence and the sharing of information objects. People are engaging in social networks in a steadily growing manner and share information objects within their communities. The high initial amount of data in such networks can serve as foundation of serious investigations towards social interactions of communities of learners. In this paper we introduce the technological foundation and architecture to model, obtain and store such user and object in- formation in so-called Artefact-Actor-Networks. Artefact-Actor-Networks combine classical social networks with artefact networks that are constructed by the use of the information objects and their connections.
  • Konferenzbeitrag
    Meta-rules: Improving Adaptation in Recommendation Systems
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Romero, Vicente; Burgos, Daniel
    Recommendation Systems are central in current applications to help the user find useful information spread in large amounts of post, videos or social networks. Most Recommendation Systems are more effective when huge amounts of user data are available in order to calculate similarities between users. Educational applications are not popular enough in order to generate large amount of data. In this context, rule-based Recommendation Systems are a better solution. Rules are in most cases written a priori by domain experts; they can offer good recommendations with even no application of usage information. However large rule-sets are hard to maintain, reengineer and adapt to user goals an preferences. Meta-rules, rules that generate rules, can generalize a rule-set providing bases for adaptation, reengineering and on the fly generation. In this paper, the authors expose the benefits of meta-rules implemented as part of a meta-rule based Recommendation System. This is an effective solution to provide a personalized recommendation to the learner, and constitutes a new approach in rule-based Recommendation Systems.
  • Konferenzbeitrag
    Social IPTV: a Survey on Chances and User-Acceptance
    (18th Intl. Workshop on Personalization and Recommendation on the Web and Beyond, 2010) Schreiber, Daniel; Abboud, Osama; Kovacevic, Sandra; Hoefer, Andreas; Strufe, Thorsten
    Incorporating Social Networking and IPTV, the two arguably fastest growing and most accepted services on the Internet today, yields strong synergies. However, it opens a very complex design space of feature combinations. Selecting features, aiming at achieving the best user acceptance, hence, is a difficult, yet vital task for the development of an integrated service. This pa- per presents the results of an initial user study conducted to gain a better understanding of the complex design space. It identifies classes of demanded, promising features and indicates that social features in IPTV services in general will be well accepted, even if they are quite immersive to the TV experience. The study was con- ducted with user groups from central Europe as well as with a group from Korea.