A large amount of learning resources is available to learners on the Web. Users of
these resources are often discouraged by the time spent in finding and assembling
relevant resources to support their learning goals, and these users often face the
information overload problem [
The interaction of previous learners with resources and the resulting
outcome can be viewed as a learning experience. An experience-based
recommendation approach would allow the experiences of similar users to be reused for
making recommendations to new users. Currently, some recommender systems
in e-commerce can capture the experience of users with items and reuse these
to enhance recommendation [
A key contribution of this research will be the development of innovative approaches to incorporate the learning experiences of previous learners captured in outcomes such as reviews and ratings, in the recommendations made to new learners. This research will harness the wide range of available e-learning resources in order to cater for learners with different preferences. The knowledge contained in the learning resources will be employed for refining learners’ goals and indexing new learning resources. This work will improve the current state of e-learning systems by reusing the experiences of previous learners when recommending relevant learning resources to new learners.
Copyright © 2015 for this paper by its authors. Copying permitted for private and academic purposes. In Proceedings of the ICCBR 2015 Workshops. Frankfurt, Germany.
This research aims to capture and reuse the learning experiences of previous learners to enhance recommendations made to new learners within a personalised e-learning system? This research seeks to address the following questions: – How can learners’ goals be refined to improve the recommendation of learning resources? – How can learners’ preferences and abilities be captured to enhance personalised recommendations? – How can learning resources be represented to support effective retrieval? – How can outcomes such as learners’ reviews and ratings, be captured and reused to enhance e-learning recommendation? 3
This research will involve the development of novel approaches for reusing the experiences of previous learners to enhance e-learning recommendation. Techniques to capture learners’ preferences and abilities will be developed. Existing learner models will be adapted for this task with the aim of capturing the preferences and the abilities of learners. This information would be used for making relevant recommendations to new learners.
Existing knowledge sources will be organised into a coherent background knowledge structure. Potential knowledge sources such as Microsoft Academic Search, the ACM Computing Classification System, and Wikipedia have already been identified. The plan is to employ these in the development of a background knowledge structure which can be employed for refining learners’ goals and for indexing learning resources. This structure will be useful for identifying the links between resources and for recommending relevant resources.
Methods for representing and refining learners’ goals will be developed. This is necessary in e-learning because learners often have insufficient knowledge of the domain to formulate suitable goals. The plan is to map the goals to a resource representation developed using shared background knowledge, this will entail reasoning with the text in the goals and the learning resources.
Representations that capture learners’ outcomes will be created. Learners’ test scores, reviews and ratings can be viewed as outcomes in an e-learning domain. Currently, learners’ test scores are the major form of feedback used in e-learning. However, this does not capture learners’ opinions which can be effectively employed to inform other learners. The plan is to incorporate useropinions with user-performance to enhance the recommendation process. 4
The research methodology has been substantially developed. Various approaches for representing learning resources have been investigated, these range from knowledge-light to knowledge-rich approaches. Some methods of refining learners’ goals have also been examined.
Different types of learning resources have been identified to use as data for this work. These include e-books, online teaching slides and video lectures. They have been chosen because they contain structure and metadata that will help with the research, and because of the variety of media types contained.
Preliminary experiments have been carried out to develop a background knowledge structure to use for the refinement of learners’ goals and the representation of new learning resources. A collection of 217 e-book chapters from the machine learning domain were collected for the experiments. Terms and phrases were extracted from the Tables-of-contents (TOCs) of the e-books using some NLP techniques and phrase identification methods.
E-books are used as the primary data source in this work because of the
structure they contain and because they are designed to be effective for teaching
and learning. Furthermore, issues of trust and provenance [
Terms and phrases were extracted from the TOCs of e-books and compared with phrases from the Machine Learning category in Wikipedia to generate a set of suitable phrases to use for developing the background knowledge structure. The result was 90 phrases consisting of 17 unigrams, 58 bigrams and 15 trigrams.
Initial output shows the potential to harness the knowledge in e-Books and Wikipedia for developing a background knowledge structure that will enable the refinement of learners’ goals and indexing of new learning resources. Further work will involve evaluation of this method, and the development of a system that employs the background structure to recommend relevant learning resources.