Unterstützung des ressourcen-basierten Lernens in Online Communities - Automatische Erstellung von Grosstaxonomien in verschiedene Sprachen
Key: Dom13-1
Author: Renato Domínguez García
Date: April 2013
Kind: @phdthesis
Abstract: Due to constantly changing professional environments and the decrease in the half-life of acquired knowledge, flexible forms of knowledge and skills acquisition are required. Nowadays, the knowledge acquired in educational institutions no longer last a lifetime. Rather, there is an increasing need (especially in work processes) for the personal acquisition of knowledge depending on specific tasks. This is called self-directed learning, as learners are responsible for their learning processes. At the same time, the World Wide Web has become one of the most important sources for knowledge acquisition. Self-directed learning using resources from the Internet is also called resource-based learning. One of the biggest challenges in resource-based learning is finding relevant web resources on the Web. Search engines are very often used for this purpose, but they do not provide assistance in the selection of found resources. Recommender systems can be helpful to find relevant resources. Learners can benefit from resources that other learners with similar knowledge requirements have already found. In larger groups or in a community, there is a high probability that relevant resources have already been found by other people. The goal of this thesis is to support resource-based learning within a community of learners by recommending knowledge resources that other community members have already found. To achieve this objective, the application scenario and an example implementation, CROKODIL, were investigated. The investigation revealed that the recommendation of interesting resources is often impossible, if the users use different terminologies for the tagging of resources. Based on this observation, a concept was developed that fills the gaps in the terminology used by the users through the use of a taxonomy. The analysis also reveals that the application scenario is characterized by current terms in multiple languages which are used as tags. A taxonomy used for the purpose of finding relationships between tags must, therefore, be characterized by the fact that it is up-to-date and available in multiple languages. These characteristics make manually created taxonomies unsuitable. Therefore, two approaches, TaxWikiHeur.KOM and TaxWikiML.KOM, were designed and implemented in order to generate large-scale taxonomies from the online encyclopedia Wikipedia in multiple languages. This is done by classifying pairs of categories from the Wikipedia in hyponymy and non-hyponymy relationships. These methods are characterized by the fact that they do not use external, manually created knowledge bases. Thus there is no need for the manual maintenance of taxonomies for new knowledge fields. TaxWikiML.KOM extends TaxWikiHeur.KOM and fixes some of the recognized shortcomings in the evaluation of TaxWikiHeur.KOM. The evaluation of the whole process has shown that even if no external knowledge base was used, the quality of the taxonomies was still very good. The approaches were evaluated in five different languages, in order to show the language-independency of the approaches. TaxWikiML.KOM was also used within CROKODIL to complement automatically generated relations between tags used by the users to describe the resources in their learning processes. Based on three corpora obtained from the application scenario, the evaluation could show that the density of the network grew using the implemented concept. Therefore, recommender systems have more information available to generate recommendations and this can be used for recommendations to learners using different terminologies. Additionally, the positive effect on the quality of recommender systems due to hyponymy relations between tags found by TaxWikiML.KOM was demonstrated in a further evaluation based on a state-of-the-Art algorithm. Finally, the FReSET tool for the evaluation of recommender systems was developed. FReSET can be used for the evaluation of recommender systems as it allows a standardized and thus comparable evaluation of recommender systems.
View Full paper (PDF) | Download Full paper (PDF)
Official URL

The documents distributed by this server have been provided by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, not withstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.