Farbod Aprin, Sven Manske, H. Ulrich Hoppe
Over the last ten years, as the digitalization increased in educational systems, the PDF format has become one of the essential document formats for lecture materials. With consideration of increasing availability of tablets and portable devices for students, web-based and shared annotations of such learning materials gained popularity. Annotating learning material is a method to promote engagement, understanding, and independence for all learners in a shared environment. OER (Open Educational Resources) have the potential to add valuable information and to close the gap between learning materials by automatically linking them. However, current popular web annotation tools for learners, such as Hypothesis and Kami, do not support learners in the discovery of new learning resources based on the context, the metadata and the content of the annotated resource. In this article, we present Salmon, a collaborative web-based annotation system, which dynamically links and recommends learning resources based on an-notations. It facilitates methods of semantic analysis in order to automatically extract relevant content from lecture materials in the form of PDF documents. Salmon categorizes documents automatically in a way that finding similar resources becomes faster for the learners and they can find communities for topics that they are interested in. We evaluated the categorization of learning materials for our application with a gold standard library created from Arxiv.org testing two methods cosine similarity algorithm and greedy string tiling. Results indicate that cosine similarity is a useful method to approximate the similarity between learning materials.