FP7 priorities for Creativity and Learning

Over the summer holidays, EU's new FP7 priorities for 2013 funding were communicated. Challenge 8. is related to learning, namely called "ICT for Creativity and Learning"

...Europe must also support national efforts to help students to learn better, teachers to teach better, and school systems to become more effective. This goal can be greatly advanced by learning systems that can adapt to effective use in a wide variety of diverse contexts.

Under 8.2 (technology enhanced  learning), they also mention Learning Analytics:

b) Learning analytics, educational data mining: tools and processes for collecting, storing, exploring and reasoning on large-scale educational data to better understand learners' knowledge, assess their progress and evaluate environments in which they learn. These tools and processes should aim at improving learning and teaching (including 21st century skills) for students and instructors.  

Learning analytics has now become such a hot topic. So, like all the others, I'm also studying the topic. Here is a link to my recent paper on learning analytics. I presented it this summer at “Open and Social Technologies for Networked Learning” in Tallinn.

Study on contexts in tracking usage and attention metadata in multilingual Technology Enhanced Learning

Just submitted the final version of the paper to a workshop on Exploitation of Usage and Attention Metadata (EUAM 09). Here is a one-pager about it and the link to the paper.

Study on contexts in tracking usage and attention metadata in multilingual Technology Enhanced Learning

“Context” is widely accepted to be important for correctly interpreting user input and for improving predictive and possibly also diagnostic models. But what is context, and how can it be measured? By measuring we mean to operationalise the construct and data gathering to provide values for the desired variables.

In this study, we consider the intersection of the areas of digital learning resource repositories, digital libraries and social tagging systems where users from a variety of countries use technology enhanced learning (TEL) offerings in a variety of languages. We consider usage and attention metadata as an example of the wider notion of context adapting the definition of context as “any information that can be used to characterise the situation of entities” [Dey01]. We give an overview of dimensions of context that are relevant in TEL, specifically arguing that context comprises the usage situation and environment as well as persistent and transient properties of the user. Therefore, distinguishing between the macro-context and the micro-context of TEL is useful.

TEL and the analysis of the data it generates take place in different types of educational settings which we call the macro-context of TEL. We use the term micro-context to denote the context that is relevant for interpreting a specific user input and for designing adequate system responses and other output. The micro-context is subdivided into user models, material/environment models, interaction models, and background knowledge, showing that usage and attention metadata are of different types and play different roles for learning about context.

We then concentrate on teachers using learning-resource repositories as an important use-case example of TEL and focus on language and country as context variables. We describe different ways in which these variables are operationalised, and we outline ways in which TEL use such context information to improve the use and reuse of repositories by supporting users in a multilingual and multicultural context. A key theme of our article is the central role that social tagging can play in this process: on the one hand, tags describe usage, attention, and other aspects of context, on the other, they can help to exploit context data towards making repositories more useful, and thus enhance the reuse.

Riina Vuorikari 1,2, Bettina Berendt3
1 European Schoolnet, Brussels, Belgium,
2 OUNL, Heerlen, Netherlands,
3 KU Leuven, Belgium

Visualising networks of learning resources

I'm looking at the first dataset of bookmarks from MELT portal. Here you can see some of the first descriptions created by using Many Eyes. Click on the interact button in the pic and it loads. This is a treemap visualisation of the bookmarks that users so far have found.

What do you see here? You first see boxes in different colours. They are "boxed" by the user IDs. The bigger one is, the more learning resources this person has bookmarked. If you hoover your mouse over the boxes, you can see the ID of resources. These, of course, do not mean nothing to you now, but imagine if they were links to resources?

Next you can explore the data a bit further. Drag the mother tongue box on the top of the graph to the first place. Now, the boxes are displayed by the languages spoken by users. You'll see that Hungarian speakers have been busy on the portal, they have the most bookmarks.

Third, you can explore further by dragging the obj_lang to the first place. This shows the languages in which the bookmarked resources are. Interestingly, it turns out, most of these resources are in English. However, the diversity is there to be observed: users have found resources in many different languages useful.

Let's go further. The next one is a network diagram. If you click on "click to interact" you can also zoom into the visualisation.

What do you see here? It's a network that consist of: user mother tongue and the learning resource that those users bookmarked on the portal. You see 4 quite big vertices, which are the mother tongues of the users.

..network consists of a set of objects called vertices connected by edges. The visualization of the network is optimized to keep strongly related items in close proximity to each other. In this way, the overall arrangement of vertices in the network is very telling of the structure of the connections between vertices (vertices that are far away are weakly related to each other).In this visualization, the size of a vertex is proportional to the number of edges emanating from it.

Take the Hungarian speakers, for example. They are the ones who user the portal most, and have actually bookmarked a fair amount of resource. At the end of each edge you can see an ID number. Those are the ID of learning resources that these teachers have bookmarked. The same goes for Finnish speakers, Dutch speakers, etc.

Interestingly, we can see from this visualisation that not many resources are shared among the users from different language groups. A few are, though: take, for example, the LeMill resource that is visualised in orange in the image here. It has edges linking it to Finnish, German and Hungarian speakers. I counted 14 resources in this small dataset that were shared by users from different countries, that's about 13% of resources.

This type of resources are what we call "travel well" resources, as they can cross borders. In this case those borders are lingual. The resource also acts as a bridge between these different language communities. If you look at the resource in question, you'll find that it is to teach English (as
foreign language) and it is in English. Thus, it is not that surprising that it is well accepted in many language communities.

Finally, I also visualised the languages of learning resources instead of the resource ID. You can find it here. As you see from the image on the right, I have highlighted the languages of resources from Dutch speaking users. They have been pretty busy finding resources in all kinds of languages!

Mine/d your data

I just participated in a week-long datamining course at the university. It was hard work, but actually a lot of fun. We plowed thorough a lot of things; including association rules, clustering, logistic regression, decision trees, neural networks, and also learned, well, made acquaintance with, some of the dataminging software like SAS Entreprise miner and used MatLab to check out the neural networks. What a strange world.

In one exercise we used the German credit dataset and wanted to come up with a decision tree to sort out the bad customers from the good ones. After lots of clicking and choosing values and setting roles, we came up with a tree that had an error rate of 47%. Wow. As well the banker could just flip a coin to choose which customer to give credit and whom not. Ok, probably a bad example, we did learn after that about the cost of misclassification, so we were able to make something better. But anyway, it just kind of made me laugh.

I was reading this blog and came across this interesting information about datamining methods that "miners" choose to use. Now that I know what all those words mean, this became an interesting piece of information for me :)

• Correspondingly, the most commonly used algorithms are regression (79 percent), decision trees (77 percent) and cluster analysis (72 percent). Again, this reflects what we have seen in our own work. Regression certainly remains the algorithm of choice for large sections of the academic community and within the financial services sector. More and more data miners, however, are using decision trees, and cluster analysis has long been the bedrock of the marketing community.

I personally thought that most useful techniques for me could be mining association rules, clustering analysis and maybe the use of decision trees. To be seen.

What I was actually pretty amazed about was that Datamining is very related to predicting missing values, i.e. the same methods that many recommender systems/studies use to predict the missing values of ratings. Another thing which was totally new was that Datamining and Machine learning are actually very related, well, quasi-overlapping, I guess.