engleski

Preporučiteljski sustavi svjesni konteksta namijenjeni autorima sadržaja u e-učenju

The goal of this thesis is to research context dimensions and characteristics of context-aware recommender systems designed for authoring of e-learning content. Building on characteristics of available content models, the context model is proposed, comprising context dimensions from (i) learning management system and course context (ii) author’s profile and (iii) context inferred from already used learning objects, which includes their structure, pedagogical roles, domain topics and layout features. Based on this model, the recommender system structure is proposed, with algorithms for analyzing content and inferring context data which can be grouped in three subtypes: (i) domain keywords and concepts, (ii) context dimensions for structure, layout and pedagogical roles, and (iii) author’s feedback. Following the analysis, data is converted to three matrices, to be used in further algorithms. Recommended learning objects are obtained from popular Web2.0 content providers according to content keywords. Received results are additionally analyzed in contextual post-filtering phase, according to relevant context dimensions and implicit author’s feedback. Implementation of model and recommender system is developed in the popular learning management system Moodle, as AREC recommender block, providing content recommendations to authors. Evaluation, based on publicly available and free content, shows that using context data improves recommendation process, which would otherwise depend only on content-based recommendation from content providers. During evaluation discussion, several open issues in model and system proposal, as well as in reusing e-learning content, are noted and discussed, with possible solutions offered.

učenje podržano tehnologijom; e-učenje; stvaranje sadržaja; preporučiteljski sustav; kontekst; sadržaj; objekt učenja; sustav za upravljanje učenjem

nije evidentirano