Content uploaded by Carlos Turro
Author content
All content in this area was uploaded by Carlos Turro
Content may be subject to copyright.
Video Learning Objects creation with Polimedia
Carlos Turro, Aristóteles Cañero and Jaime Busquets
Área de Sistemas de información y comunicaciones
Universidad Politécnica de Valencia
Camino Vera s/n, 46022, Valencia, Spain
{turro, acanero, busquets} @asic.upv.es
Abstract— In this paper we present Polimedia as an alternate
way of producing video learning objects instead of lecture
recording. We are using heavily this technology with a teacher
support program involving more than two hundred teachers
and in this time we have produced more than two thousand
Polimedia learning objects and we have also disseminated this
technology with several universities. We present also results of
using this technology together the support program.
Keywords: learning objects, lecture recording.
I. INTRODUCTION
Lately there has been much work around lecture
recording systems. On those systems a more or less complex
set up is prepared to record simultaneously audio, video and
teacher computer’s screen in order to get a high fidelity view
of a lecture.
The presentation of those recorded lectures and other
talks takes various forms: Simple combinations of audio and
presentation, combination with presenters‘ video (“talking
head”), picture-in-picture with optional re-sizing and
highlighting of the two elements (video and presentation) or
a synchronized side by side – the options are manifold.
While they are able to convey the content of the recording,
their visual appeal is limited when compared to customized
productions or TV-style webcasts.
There are many examples of lecture recording systems.
From late 90’s University of California Berkeley [1] has
been recording live sessions. Also there are many lecture
recordings into the OCW project [2]. A common issue in
these projects is the need of making an automated production
as a requirement for scaling lecture recording for a massive
deployment,. There are some works on this field, like in [3]
[4]. Also there is a need for indexing that lectures. That can
be done through automated speech processing, like in [5].
Lately, ETH Zurich’s Replay [6] and VirtPresenter [7] are
good examples of the capabilities of those lecture recording
systems, and the recent Matterhorn [8] project from the
Opencast consortium [9] is filling the gap for an enterprise-
level platform for lecture recording. .
At Universidad Politécnica de Valencia (UPV) we
decided in 2003 to take a slight different route for video
content creation. We record small pieces of synchronized
screen + audio + video in a specialized production studio and
we add a pedagogical content to create a learning object, that
will be distributed offline through a Learning Management
System, in our case Sakai [10]. The fact that the teacher is
not in a lecture hall and that he has no real students in the
recording allows some interesting possibilities. The process
of creating those video learning objects is what we call
Polimedia.
The idea behind Polimedia is that when you record a
lecture in a standard hall, you have a content that is not
atomic in the sense that is formed of diverse topics, and is
not easily reusable because that content refers to other topics,
usually from previous lectures. For instance, in Berkeley’s
webcasts [11], each year has its own recordings for the same
topic. Instead of this approach we decided to create Learning
Objects (LO), small atomic pieces of reusable content, and as
this creation can’t be done easily on a lecture recording hall
we use a recording studio, that allows us to improve the
recording quality. Finally a teacher can use any LO in a
lecture session, or have it as a learning material.
Also we realized that there is also a need for quality
reviewing in the content production scheme. At UPV we
designed, and we have used for three years a complete
program of content production, beginning with teacher
selection, content creation, quality assurance and rights
management. That program has allowed us to have currently
a portfolio of 4535 Polimedia objects from 734 different
teachers, with a total time of 698 hours.
The rest of the paper is organized as follows: in section 2
we will explain the Polimedia approach to video content
production. In section 3 we will present the learning objects
paradigm. Section 4 will be devoted to the digital production
program. Section 5 deals about the results we have obtained
from Polimedia and the digital production program. Finally
in section 6 we will draw some conclusions and we will talk
about our planned future work and interaction with external
projects.
II. THE POLIMEDIA PRODUCTION SYSTEM
Any lecture recording system aims to produce a high
quality and visually appealing view of the teacher, the
computer screen or the blackboard, and the audio recording.
That view has to be clear, smooth and must not distract the
student from the learning process. With these issues in mind
we have stacked horizontally the slide and the presenter,
composing a 1088x672 frame, as you can see in figure 1.
If we show only the bust of the presenter we allow a little
overlap between the presenter and the slide, maintaining the
frame size. It is only noted when the presenter extends his
right arm. This is depicted on fig. 2.
Figure 1. Polimedia full shot
We have designed the process carefully to achieve both a
high rate of production and to get a final quality is
comparable to a TV production, but with much lower cost. A
key piece in this setup is the production studio.
A Polimedia production studio is a 4x4 meters room in
which we deploy a camera, two PCs, a pocket microphone,
lights and some A/V equipment, including a video mixer,
and an audio noise gate. It is worth noting that the use of
lights in such a reduced set allows getting a sharper image
easier than in a lecture recording hall. The cost of this
studio is around 30.000€.
When a teacher wants to record a Polimedia, he/she
arrives to the studio with the slides and shows a presentation
directly on the studio. There we record both the screen and
the video from the teacher in two different streams.
After that we have a raw preview of the Polimedia
content by putting the screen and the teacher video side-by-
side, and on that screen the teacher can review the material
he has just created. If everything is OK we start a process for
cropping, joining with a little overlap and encoding, in order
to have an .mp4 file with h.264 encoding, suitable for
distributing and archiving.
Then, the cropping, joining and encoding process is fully
automated using avisynth[6] and ffmpeg, so the teacher can
see a preview of the recording immediately and the resulting
file is ready in a few minutes. As stated, the final file is a
.mp4 file at 500kbps, that we distribute by streaming through
a flash media server. Also we encode an mp4 designed for
mobile devices, like IPhones.
Figure 2. Polimedia half shot
From teacher’s point of view, recording a Polimedia
object is as easy as arrive with a PowerPoint file, a laptop, or
even a URL to our recording studio. There are two screens,
on at the front of the teacher, and one at the right. In such
position he or she will record the lesson and both streams
will be recorded. In figure 3 you can see the set up for the
studio and an image of a live recording. Actually we have 8
Polimedia production studios.
Polimedia content is by design asynchronous, so is
intended to be distributed through a LMS or a website. We
think the best way of knowing what Polimedia can do is by
viewing some examples of them. Here are the direct links to
some that represent the possibilities of our work. Please
remember that you can access more examples at
http://polimedia.upv.es/polimedia or at our Library Dspace
repository at http://dspace.upv.es :
• http://polimedia.upv.es/visor/?id=32d880f6-709a-
d74b-b7ad-b4394d9ccbfd
• http://polimedia.upv.es/visor/?id=cc42af48-5c15-
684c-9e31-53eb2b7c2581
• http://polimedia.upv.es/visor/?id=59436990-49f6-
e64d-b2de-1598776a845b
• http://polimedia.upv.es/visor/?id=98cf78f3-aa20-
7142-bf3e-e8f19d6b1efb
So, the main characteristics of the Polimedia tool can be
summarized as follows:
• Polimedia allows fast and affordable creation of digital
contents. The cost of a study of production does not surpass
in any case the 30.000 €, and is operated by just one
technician.
• Polimedia does not require special knowledge by part
of the professors or expert.
• Polimedia produces high audiovisual and didactic
quality content, adapted to its distribution through Internet
with any format by transcoding.
In year 2008 we began to disseminate Polimedia
technology to other Universities and e-learning projects.
Currently we have deployed Polimedia at 10 Spanish
Universities, and we are focusing on getting 4 more before
the end of this year.
Also we have contacted with some European
Universities, like Aarhus at Denmark, and we are willing to
support deployment on those Universities. Finally we are
putting Polimedia on the AVICENA project, funded by the
UNESCO, willing to give e-learning through all African
countries and we have installed a Polimedia studio in Dakar.
Anyway, we are looking for partners willing to work
together with us in further improvements of Polimedia
technology.
Figure 3. A Polimedia recording session
Also we can give feedback about how Polimedia is used
on Latin America. The Universidad Politécnica de Valencia
uses Polimedia to provide e-learning to that continent, and
we have a strong tie with Mexico, where we are fulfilling
successfully some e-learning needs.
III. DIDACTICAL APPROACH. LEARNING OBJECTS AND
POLIMEDIA
The Polimedia system belongs to asynchronous training
e-learning category using the Learning Objects (LO)
paradigm. As defined by Willey[7], a LO is the minimum
unit of digital learning content that can be reused and
sequenced. So these small components should work like
integrated circuits of the teaching-learning process, offering
the students the possibility to improve their performance and
level of satisfaction.
Then, the determination of the features of the learning
objects facilitates the establishment of some appropriate
approaches of their validation, and at the same time it will
help in assuring the quality in their creation. So a LO should
have:
1) Digital format: A LO should be usable from Internet
and accessible simultaneously for many people and from
different places.
2) Pedagogic purpose: the objective is to assure a process
of satisfactory learning. Therefore, LO will not only include
contents but rather also guide the own process of the
student's learning.
3) Interactive content: A LO should encourage as much
as possible the active participation in the information
interchange. Then a LO should tend to include activities
(exercises, simulations, questionnaires, charts, graphics,
slides, tables, exams, experiments, etc.) that allow to
facilitate the process of assimilation and the tracking of each
student's progress.
4) Indivisible and independent of other learning objects,
so we can make different learning sequences for one LO,
depending on the student or the objectives.
5) Reusable in educational contexts different to that for
the one that was created. This feature is the one that
determines that an object has value, being one of the starts
that base the concept of learning object. For a LO can be
reusable it is necessary that there isn’t any reference to their
location in any particular course neither in the subject nor in
the course. Also a LO will use metadata to describe its
content.
6) Synergy: Learning modules can be created combining
different learning objects within a certain context
By using the LO paradigm, we structure the course in
modules, and then in lessons, as seen on next figure. Each
lesson will address just one concept, so that we can reuse the
content. For instance in next figure we have some Polimedia
learning objects, that we could arrange in different learning
sequences. In a more general case we access learning
modules composed of learning objects through or Sakai
learning platform.
In general terms, the learning objects paradigm suppose,
on one hand, a saving of time for the professor in the
preparation of resources of learning of quality, and on the
other hand, a constant update of the resources for the student.
When creating a Polimedia, any teacher has to deal with
some requirements in order to achieve a successful work.
The main ones are:
- Structure the course in learning objects
- Use a clear, simple language, close to the listener.
- Take care as much the verbal communication as the
non-verbal communication.
- Have a maximum length of around 10 minutes. We
have carried out a study with all our Polimedia
objects that reinforce the believing that when the
learning object is larger, the less is completely
viewed.
While a course can be made entirely of Polimedia LOs,
that can be purposely recorded or reused, it is true that there
is learning content in other formats (text, quizzes, etc.) that is
useful also as a different kind of LOs that can be mixed
altogether. After that there is also information about the
course, a summary of a lecture and other material that does
not satisfy the requirements of a LO but there is used within
a lecture. We call that material coupling objects and are used
in the creation of a learning module, as we will explain in
section 4.
IV. ENCOURAGING LEARNING OBJECTS PRODUCTION.
In the previous section we defined a Learning Object
(LO) as a digital, self-contained, reusable entity with a
learning aim that contains the content, the metadata and the
information which helps its identification, storage, and
recovery.
Creation of these objects is not as easy as in the Lecture
Recording scenario; in with we try to extract usable content
from standard teaching. In our paradigm we want to design
the content to have the properties that define a LO, and that
requires a previous work. For supporting teacher’s work the
Universidad Politécnica de Valencia has a support program
named “docencia en red” that gives both technical and
economic support to teachers that create learning objects.
To solve that problem we have designed a process that
helps teacher to get good results from their valuable time.
These are the steps that our teachers follow in the creation of
LO.
1. Determine the objectives: conceptual, procedural or
attitudinal
2. Create the objects
3. Choose format (text, multimedia, interactive)
4. Create the introduction of the content
5. Create the content
6. Fill up the online metadata form. We use the
standard LOM tags.
7. Specify access rights
8. The LO Comittee evaluate the quality of the
learning object. The LO committee is formed by teachers
and pedagogy specialists, appointed by the dean and make a
review focusing on the formal aspects of the content and the
overall result. Its approval is required for a teacher to get the
economic incentive assigned to a LO creation.
After the LO creation it time to create Learning Modules.
A Learning Module, as shown in figure 4, is created by the
contextualization of one or several learning objects within a
didactic environment with the addition of coupling objects
The approximate working time of the student in a
learning module is one hour, including all the activities
proposed in the module
So, the detailed process of creating learning modules is:
1. Select the competences to develop in the student
2. Select the kind of objects to work in: conceptual,
procedural or attitudinal
3. Define the structure of the module
4. Search for already created learning objects which
can fit in our structure
5. Create non-existing learning objects
6. Configure the module: create coupling objects
7. Test the module
8. Evaluate the module by a LO Committee
All the objects are stored in a common repository,
catalogued by the University Library in a DSpace repository.
Currently we have two types of access rights: Any student
from the UPV and Free access with a Creative Commons-
like license. Even that we prefer the second one, the kind of
license is not determined by the plan and each teacher can
decide which one to use.
Figure 4. Process of creation of a learning object
Figure 5. Polimedia objects viewed by year
It is worth noting that as we have a LO Committee that
evaluates the quality of the objects, not all our content has
been approved, mostly because of the requirement for a LO
to be self-contained. That content can be useful too, and
teachers are allowed to use in their lectures.
V. STUDYING POLIMEDIA AND LO USAGE. COST ANALYSIS
A. Polimedia recordings
Polimedia operative recordings began in UPV at year
2006, and we have tested or setup for all these years. We
have recorded 4535 Polimedia objects from 734 different
teachers, with a total time of 698 hours. That content has
been viewed about 370.000 times, from 95 different
countries. This is shown on fig. 5.
We have carried out study in which we analyzed the log
from our Polimedia streaming servers to know about the
abandon rate, that is how long users see Polimedia LO. This
study has been made with all accesses to our servers in the
years 2008 and 2009, with more than 150.000 hits.
First we look at the percentage of time users see our
videos, that is depicted in figure 6. In this graph we represent
on the X-axis the percentage of viewing time, that is if a user
see fully a LO that time is 100% and if he sees half of time it
is 50%. It should be noted that we use a streaming server that
provides that kind of statistics, because if anybody does that
with a progressive download web server (like Youtube) you
don’t have such capabilities.
Figure 6. Polimedia viewing time
Figure 7. Learning objects mean viewing time
Viewing figure 6 we can identify three user profiles:
• Those that open the object and abandon immediately,
generally because it was not the object that wanted to see.
This corresponds to errors, tries and similar elements, and
corresponds roughly to a 20% of the cases.
• Those that see entirely the video (x-axis 90-100%).
Here we have 48% of the users and a 60% if we deleted the
percentage of errors.
• Those that effectively drops, with a time declining
profile.
In figure 7 we compare these data with the length of the
LO, that is what is the mean abandon rate for videos with
length x. As you can see, the longer the LO the higher the
abandon rate. So, it is better to design several shorter
learning objects than one longer one. This results support our
recommendation that a Polimedia object should have a
maximum length of around 10 minutes.
B. Learning Objects support program
The “docencia en red” program began supporting
teachers to create LOs in 2007, and we added support for
learning modules in 2008. Since then the program is now in
its fourth year, and we have created 2451 Polimedia learning
objects and a total of 4872 LOs, as shown on fig. 8.
C. Cost Analysis
In the introduction we face the LO & Polimedia method
of content creation with the lecture recording approach.
While the objective of this paper is not to compare both
paradigms, it is true that it is necessary to take account on the
costs of a Polimedia recording.
When we appoint a Polimedia studio we reserve about 60
minutes for the recording of two Polimedia, or twenty
minutes of content, so we take into account that we need
three times the teacher’s time for recording.
There is also a need for restructuring the content on the
LO paradigm. This depends very much on the subject and
teacher, but we usually estimate this time in one time more,
so finally we need four units of teacher’s time to produce one
unit of Polimedia content.
Figure 8. Results of the plan
There is also a need for a technician on the studio. As the
production scheme is automated, we will need one unit of
technician time for a unit of content.
So creating a Polimedia costs about four times more than
a lecture recording and should be used at least four times
more.
VI. CONCLUSIONS. FUTURE WORK
In this paper we have presented Polimedia, a way to
produce, video e-learning content with both a high quality
and a minimum work for teachers. Polimedia technology has
been heavily tested, not only for the technological side, but
for the pedagogical side too. Polimedia content creation is
tied with the learning objects paradigm, because we have to
take teachers out from the lecture hall to a recording studio,
and structure the lecture in a different way.
So we developed, and we have also presented the
“docencia en red” program that support teachers in using this
technology. The program has been very successful and we
are creating more learning objects and modules every year
and we have presented too some numbers on the use of
Polimedia learning objects by the students.
On the other hand, the cost effectiveness of our scheme
depends heavily on the reuse of the LOs. If they are never
reused we have wasted teacher’s time and University’s
resources. That reusability involves more resources than just
the objects. For instance, without good metadata a teacher
cannot find what LOs have been created and without a clear
view of rights can’t decide if the content may be used.
Now we are working on joining efforts with the Opencast
consortium and we aim to develop an extension for
Matterhorn to include Polimedia creation into that open
source project. We are advancing very fast and hopefully we
will get results soon.
REFERENCES
[1] Rowe, L.A., Harley, D., Pletcher, P., Lawrence, S. (2001). BIBS: A
Lecture Webcasting System. BMRC Tech. Report June 2001, Univ.
California, Berkeley.
[2] http://ocw.mit.edu
[3] Machnicki, E. and Rowe, L.A. (2002). Virtual Director: Automating a
Webcast. Proc. SPIE Multimedia Computing and Networking,
January 2002.
[4] Bianchi, Michael: Automatic video production of lectures using an
intelligent and aware environment. In: MUM ’04: Proceedings of the
3rd international conference on Mobile and ubiquitous multimedia,
117–123, New York, NY, USA, 2004. ACM.
[5] Glass, James, Timothy J. Hazen, Scott Cyphers, Igor Malioutov,
David Huynh und Regina Barzilay: Recent Progress in the MIT
Spoken Lecture Processing Project. In: Interspeech 2007, pp. 2553–
2556, August 2007.
[6] Schulte, O. A., Wunden, T. and Brunner, A., “REPLAY - An
Integrated and Open Solution to Produce, Handle, and Distribute
Audio-visual (Lecture) Recordings,” Proceedings of the 36th annual
ACM SIGUCCS conference on User services conference, Portland,
Oregon, October 2008.
[7] Mertens, R., Ketterl, M. and Vornberger, O., “The virtPresenter
lecture recording system: Automated production of web lectures with
interactive content overviews,” International Journal of Interactive
Technology and Smart Education (ITSE), 4(1), February 2007.
Troubador publishing, UK, 2007, pp. 55-66.
[8] Markus Ketterl, Olaf A. Schulte, Adam Hochman, "Opencast
Matterhorn: A Community-Driven Open Source Solution for
Creation, Management and Distribution of Audio and Video in
Academia," ism, pp.687-692, 2009 11th IEEE International
Symposium on Multimedia, 2009
[9] http://www.opencast.org
[10] Sakai (2008). Sakai Success Story. Retrieved from
http://www.sakaiproject.org/access/content/group/sakai-
community/CaseStudies/StanfordCaseStudy.pdf
[11] http://webcast.berkeley.edu/
[12] http://avisynth.org
[13] Wiley, D. (2002). The Instructional Use of Learning Objects.
Retrieved from http://www.reusability.org/read