The Use of User-based Video Tags in a Recommender System Scenario for Learning Material prediction

Benedikt Engelbert, Karsten Morisse, Oliver Vornberger
Hochschule Osnabrück, Universität Osnabrück

The Internet is a network, where data and information can be accessed immediately.
Also in the area of e-Learning it is more and more common, that digital learning
material is accessable over the Internet, whereby the access is easy and speed up
the distribution. Through this, we have a huge variety of learning material, since the
student can not only choose between script and slides, but also between audio-
/video content, online tests and more digital material. This seems to be an advantage
at first, but the variety can result in disorganization, excessive demand or the
ignorance of unused material. Those circumstances are fairly common in scenarios,
where loads of information is provided. Therefore, many Internet services provide
assistive systems called recommender systems. Recommender systems help users
to find the most suitable or important information regarding his or her interests. As
one of the most popular services the online store of can be named,
where users get recommendations depending on already bought items. So, if a user
bought an item like a book about video streaming technologies, the user will get
recommendations for further literature in video streaming and similar topics. Such
recommendation systems are also necessary in e-Learning environments to
counteract the circumstances contingent upon excessive demand and to enhance
the learning process of students. Based on this, we’re working on a recommendation
system to predict learning material regarding the current needs of a student. The
assumption is, that a student’s learning process can be enhanced, if the following
factors are respected:

  • Appreciate student’s cognitive ability
  • Recommend unused learning material
  • Recommend the most appropriate learning material regarding the student’s
  • current performance level
  • Support in organizing learning material

Our approach to achieve a suitable data base for recommendation process is a tag
based learning scenario. We provide a learning platform with the possibility to work
on provided learning material online regardless of whether it is a PDF document,
power point slides, audio-/video content or something else. The students can add
tags to a certain content he or she is working on. A tag is at least a label, where the
student can add additional information for the material or for a part of the material. It
is necessary to outline, that a tag can be added to a certain area of the material too.
To picture this, the idea for a text document is shown in Figure 1, where three tags
were added to areas in the document.

Two tags provide new materials, which the student added. The other tag provides a
comment. It is also possible to add a rating for a certain area or page. All those tags
are stored in a user profile called learning profile. The system provides a learning
profile for each student. While he or she adds tags to the material he or she is in a
certain state of the lecture topic. The lecturer needs to add those states before. The
easiest case could be the chapters of given lecture notes. The real benefit comes up
with the aggregation of all profiles. As pointed out in Figure 2 the aggregation results
in “hotspots”.

It can be assumed, if a student has a problem with understanding a topic that there
are other students with the same problem. In this case (cp. Figure 2) there is material
from another student available, which could be a benefit for the others who have
worked with the text document. The system interprets those detected “hotspots” and
recommends appropriate content.
Tagging in a static document is fairly easy. It seems to be more difficult in moving
image material. In this article we’re focusing on working with moving image material
like lecture recordings or pod casts and also on the recommendation process for
those learning materials. We will point out how to use tags in audio-/video content,
find hotspot areas and use the given data base for further recommendations.
Furthermore, the differences between moving image content and static material, but
also the difficulties working with such content will be pointed out. Since the idea is not
just to recommend whole learning material, but also segmented parts of a material, it
is worth mentioning how to merge several hotspots to a topic related area.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>