In this paper, we briefly describe the goal and activities of the LAK15 workshop on Visual Aspects of Learning analytics.

This paper presents a suite for configuring and deploying mobile, location-based educational games for natural history museums. We propose a novel approach that uses structured museum-data as game-content and present an initial study thereof. The overall aim is to enable domain-experts, like teachers or museum educators, to create educational games for off-the-shelf mobile phones without requiring neither bespoke programming skills nor expensive setups. Our approach builds on the combination of two strands of work: 1.) structuring and providing museum-assets as linked data, and 2.) template-based content creation for mobiles games. This is thought to facilitate reuse of existing data and foster its maintenance on the one hand, as well as providing a timely and engaging mobile interface to museum content on the other hand. Our study is based on two rounds of user-tests with think-aloud observations and questionnaires, and on the developers' notes and reflections that stem from introductory workshops at four natural history museums from the Natural Europe project consortium. We found that both, the workflow for content-creation and the mobile end-user app have high hedonic as well as pragmatic qualities. Thus, the test results indicate that our approach might provide guidance for future work in this domain.

In order to satisfy and positively surprise the users, a recommender system needs to recommend items the users will like and most probably would not have found on their own. This requires the recommender system to recommend a broader range of items including niche items as well. Such an approach also support online-stores that often offer more items than traditional stores and need recommender systems to enable users to find the not so popular items as well. However, popular items that hold a lot of usage data are more easy to recommend and, thus, niche items are often excluded from the recommendations. In this paper, we propose a new collaborative filtering approach that is based on the items' usage contexts. The approach increases the rating predictions for niche items with fewer usage data available and improves the aggragate diversity of the recommendations.

Recent work has studied real-life social and usage datasets from educational applications, highlighting the opportunity to combine or merge them. It is expected that being able to put together different datasets from various applications will make it possible to support learning analytics of a much larger scale and across different contexts. We examine how this can be achieved from a practical perspective by carrying out a study that focuses on three real datasets. More specifically, we combine social data that has been collected from the users of three learning portals and reflect on how they should be handled. We start by studying the data types and formats that these portals use to represent and store social and usage data. Then we develop crosswalks between the different schemas, so that merged versions of the source datasets may be created. The results of this bottom-up, hands-on investigation reveal several interesting issues that need to be overcome before aggregated sets of social and usage data can be actually used to support learning analytics research or services.

Positive emotions have been proven to be a key factor for successful learning. In modern personalized learning environments informal learning takes a prominent role and with this the use of computer-mediated communication. Communication data, like for example chat logs, can be harvested for sentiments. Most sentiment analyses operate processing only verbal information. But the messages exchanged in text based chat also contain low key nonverbal communication. Emoticons are used to compensate the missing visual information and communicate emotions. In this paper we want to introduce a method to perform a sentiment analysis on text based chat logs relying only on emoticons to detect positive sentiment. We implemented and tested this method for the instant messenger Skype. Our first results show emoticons do indeed represent a strong indicator for detecting positivity within chat communication.

In this paper, we introduce a new way of detecting semantic similarities between learning objects by analyzing their usage in a web portal. Our approach does not rely on the content of the learning objects or on the relations between the users and the learning objects but on usage-based relations between the objects themselves. The technique we apply for calculating higher order co-occurrences to create semantically homogenous clusters of data objects is taken from corpus driven lexicology where it is used to cluster words. We expect the members of a higher order co-occurrence class to be similar according to their content and present the evaluations of that assumption using two teaching and learning systems.

The growing amount of available information on the internet makes the process of filtering appropriate information an increasing challenge. Because currently existing approaches provide insufficient results in many cases, we propose a new way of relating objects based on their usage. We assume that objects which are significantly often used in the same session are semantically related. Thus, we build a usage-based relatedness graph, apply a graph-based clustering algorithm and evaluate the results with respect to semantic similarity measures. Our approach takes the learning domain into special consideration, its evaluation is performed within the Learning Object Repository MACE.

We present new ways of detecting semantic relations between learning resources, e. g. for recommendations, by only taking their usage but not their content into account. We take concepts used in linguistic lexicology and transfer them from their original field of application, i. e. sequences of words, to the analysis of sequences of resources extracted from user activities. In this paper we describe three initial experiments, their evaluation and further work.

In the world of recommender systems, it is a common practice to use public available datasets from different application environments (e.g. MovieLens, Book-Crossing, or Each-Movie) in order to evaluate recommendation algorithms. These datasets are used as benchmarks to develop new recommendation algorithms and to compare them to other algorithms in given settings. In this paper, we explore datasets that capture learner interactions with tools and resources. We use the datasets to evaluate and compare the performance of different recommendation algorithms for learning. We present an experimental comparison of the accuracy of several collaborative filtering algorithms applied to these TEL datasets and elaborate on implicit relevance data, such as downloads and tags, that can be used to improve the performance of recommendation algorithms.

The generation of Contextualized Attention Metadata (CAM) allows to retrieve information about the different actions that users execute over different resources in a specific context. This paper presents how CAM is used within a learning system to personalize help provided to students while working on online exercises. We outline our approach and present two application examples within this framework for the personalization of exercises with hints.

In the MACE project we aim to support architecture students while searching for the right learning materials by connecting several repositories containing architectural learning objects and by linking these objects beyond repository boundaries. We use real world object representations which serve as connection between learning materials to allow the students the exploration of new and more complete learning paths. In this paper we outline the generation and usage of real world object representations within the MACE system and evaluate our approach.

A learner's interaction with her computer can be recorded and stored in the format of Contextualized Attention Metadata. The collected data can then be analyzed to support the learner in her self-reflection processes. We present two ways to discover patterns in the collected attention metadata by applying methodologies based on the Rough Set Theory and explain how these results can support a learner when learning in a self-regulated way.

Within the MACE project, we support simple access to and retrieval of digital learning resources in exploratory architectural education. In order to facilitate the improved access, we introduce a digital representation of real world objects through an extension of the LOM based MACE application profile, named entity recognition and the combination of various databases. Automatically grouping learning resources, we add additional metadata otherwise not available and thus improve retrieval of suitable learning resources. Our first results clearly indicate the suitability of our approach.

Today, users do not only access learning resources stored in traditional repositories. They use any kind of digital resource, often generated in the context of social media platforms. When trying to improve the learning experience, it is crucial that all resources that are potentially useful are accessible; and that the resources are described in a way that allows to identify what is relevant. We suggest enriching learning resource descriptions by combining metadata generated by experts, user interactions, observation components and machine-generated metadata. We give a brief example of the compilation of rich metadata sets by integrating the social media platform ALOE with resources and metadata from the Ariadne knowledge pool and public FlashMeeting replays.

In this paper we propose a schema and framework for recording and managing attention metadata. This framework is intended to capture, manage, and re-use data about attention users give to learning objects in different applications. Capturing attention allows us to track what objects people use and how they use them.

RDF-based P2P networks have a number of advantages compared with simpler P2P networks such as Napster, Gnutella or with approaches based on distributed indices such as CAN and CHORD. RDF-based P2P networks allow complex and extendable descriptions of resources instead of fixed and limited ones, and they provide complex query facilities against these metadata instead of simple keyword-based searches.In previous papers, we have described the Edutella infrastructure and different kinds of Edutella peers implementing such an RDF-based P2P network. In this paper we will discuss these RDF-based P2P networks as a specific example of a new type of P2P networks, schema-based P2P networks, and describe the use of super-peer based topologies for these networks. Super-peer based networks can provide better scalability than broadcast based networks, and do provide perfect support for inhomogeneous schema-based networks, which support different metadata schemas and ontologies (crucial for the Semantic Web). Furthermore, as we will show in this paper, they are able to support sophisticated routing and clustering strategies based on the metadata schemas, attributes and ontologies used. Especially helpful in this context is the RDF functionality to uniquely identify schemas, attributes and ontologies. The resulting routing indices can be built using dynamic frequency counting algorithms and support local mediation and transformation rules, and we will sketch some first ideas for implementing these advanced functionalities as well.