t-learning

Linking it all June 4, 2008

To finish with this topic, today I’ll present the solution proposed by Rey et al., a mix of what I’ve already talked about and more…

First, the user receives the subject’s educative contents and ontology (conforming ADL SCORM) from the contents provider, and the DiTV programs (conforming TV-Anytime).

As you could read previously, the programs should be tagged by the viewers, improving in this way the semantic tagging. This means that the user should be given an interface in which they could write the tags. The developed interface is seen over the program screen and it offers some facilities as suggestions to avoid writing the whole word (we must keep in mind that the user has only the TV remote control to manage the interface), previous assigned tags and popular tags for that program, as you can see in this screenshot from the article:

Now, the big question is: how can the system use the tags to provide programs related to a topic?

As the number of repetitions in global assigned tags grows, the relationships among them are computed, and also the relationships among the programs assigned popular tags. By means of these relationships, a main subset of programs can be built up, being the relevant tags those appearing in the ontology. Related programs can be added to this set, if they have been assigned some (though not all) of the relevant tags, enhacing the experience of the user. Popular tags are also sent to the provider, so that they can be delivered together with the program.

As the computation of relationships is too heavy for a simple device as a set-top box, it is done in a remote server, leaving for the device the only task of filtering the programs according to the set of relevant tags.
This server + local device system runs AVATAR, and you can read more about it in the International Journal of Pattern Recognition and Artificial Intelligence, Special Issue on Personalization Techniques for Recommender Systems and Intelligen User Interfaces, 21(2), pages from 397 to 422.

So… that’s all! I hope you enjoyed the reading and learned a bit about t-learning and this proposal :)

Folksonomy May 27, 2008

In the previous post I presented the use of folksonomy for tagging the DiTV programs, so that they could be automatically and semantically matched to a course ontology. However, it wasn’t very clear what it is and why it is useful, so today I’ll give a brief explanation on the subject.

A folksonomy consists of a series of pieces of information of any kind classified by users with labels. The labels space is plain – without hierarchy, in contrast with an ontology or taxonomy – and common among all the users. So, the term would explain this “classification made by people”, as marked by Quintarelli, though the term is attributed to Van der Wal and made popular by Gene Smith. Van der Wal defines it like this:

Folksonomy is the result of personal free tagging of information and objects (anything with a URL) for one’s own retrival. The tagging is done in a social environment (usually shared and open to others). Folksonomy is created from the act of tagging by the person consuming the information.

So, this proccess has two dimensions or steps – personal and social. First, the user marks a piece of information to have an easier access to it later. This kind of selfish behaviour would be the personomy. Then, and more interesting for the rest of the world, it also contributes to a wider classification of articles – the folksonomy.

It’s worth noting that there are two kinds of indexation made by humans – the tagging done by the creator and the one done by the community of viewers. Examples of the first are YouTube (videos), Flickr (pictures), or Tagzania (places). Examples of the second, which would be the more accurate folksonomy, are mainly bookmarking sites as del.icio.us or ma.gnolia. Now we are interested in the second type.

What do we achieve by doing this? It is clear that ambiguation and synonims may give some problems. Selfish tags like “important” or “mine”, that only helps a user to find their marked items, don’t give any benefit to the community. However, the greatest amount are describing tags that give an average view of what we can find in that item.

Who said that ‘Heroes’ had anything to do with cheerleading? May 26, 2008

In the previous post, I was talking about the need of ontologies to define the area of interest of the course, so that we could afterwards match it with the metainformation attached to TV programs. So, here we come to the conclusion that TV programs should be semantically annotated, too.

But who performs such annotation?

It would be nice that the same people designing the course also annotated the programs. This, though, is impractical and maybe impossible, and the original idea is to achieve the matching automatically, so that new programs can be added to the course without new revision.

The simplest way would consist in letting the TV channels include this information by themselves. But here we face a potential problem: TV providers might not worry about t-learning, spending little time and effort in doing this correctly – we could find wrong tags, scarce or even no information at all, worse than using syntactical search.

The approach that Rey et al. propose is the use of social tagging, and it is, in fact, the main contribution of the article of discussion for this blog. Being the actual viewers the ones that assign tags to the programs, the most popular tags for a specific one should be highly accurate.

In the next posts we’ll have a look on social tagging, which is called folksonomy, and how it can be applied to t-learning in practice.

What should I watch to learn about first aids? … ‘Blossom’? ‘ER’? May 24, 2008

3, 2, 1… Restarting… May 20, 2008

Lately my blog was in forced vacation, now it’s back! (though comments have sadly disappeared).

Please keep tuned in to know more about t-learning and Rey’s approach on the subject.