Friday, June 13, 2008

Pipe trial for delicious

Monday, June 02, 2008

This is it! Resources that cross boundaries

Ok, I think this graph is the coolest kid in the blog!!




What you can see here are the communities of users by mother tongue (nodes) and the edges are the resources that these users have added to their collections.

This is a great visualisation of communities of practice. What you can see here at a glimpse is that the learning resources that these users have added to their collections, are very much community oriented, in this divided by languages.

I sometimes frame my research question as the following:
Does a multi-lingual and multi-cultural learning resources portal rather act as one system divided into different language or country groups, or is it more like one monolingual system with its own sub-groups and communities of practice (think of a system like delicious) that cross the language and cultural borders?
This visualisation seems to point more to the first one (this REALLY needs to be further investigated!!), it seems that users are divided into groups by mother tongue. Why I say so is that you cannot see many resources that are shared among the groups.

To play around with this by yourself, make sure that you click on the arrow head down at the menu bar. This allows you to see in which directions the links go. They often time just go to one direction.

There are some resource that indicate communities of interests between countries. For example, in this image, we can see that there are some resources that are shared by both Estonian and Lithuanians. One of them is highlighted in orange.

These are the interesting resources as they cross between boundaries. The more I think of it, the more I'm convinced that you cannot call these call boundary objects (see my previous post). If I got the boundary object right, they are the objects that help these two groups to talk to one another, because they do not share the same language or jargon. But in this case, I think it's the contrary, these people share so much the same, that they can even share resources in Russian (of course being ex-Soviet countries, Russian is a common knowledge).

Anyway, even if the rather disappointing news were that users on an international portal seem to stick to one another based on their mother tongue rather than common educational interests, the good news is that I believe that through making more social cues and traces available to them, they would actually start exploring the resources in other languages and other areas.

And besides, who says that my data here really actually displays this community correctly!? This is based only on the common resources that users have put to their collections. Actually, LeMill is more of an authoring environment, so maybe a better way to study this community would be through collaborative authoring of learning resources? Or something else, like common search terms or tags that are used.

So, take this exploratory description of this data set with a little bit of skepticism!

In what languages are the resources that end-up in collections?

Well then, I guess that will be a no-brainer...

In this visualisation, you can see the languages of resources (e.g. English) as nodes and the languages of users as edges (e.g. en, de..).


If you click, for example, on English, lot of edges are highlighted. Those are the mother tongues of users who have bookmarked these resources. After little bit of playing, you'll find that English resources, and the ones with no languages, seem to be most popular with users.

However, it is cool to see that resources in other languages also end up in users' collections. Here, for example, you can see that Czech (sorry for misspelling) are used also by users with Polish and Lithuanian as mother tongue.

More analyses are needed to give you any numbers, but this already is an interesting insight.

Resources country of origin and user mother tongue

This visualisation shows the links between the country, where the resources in the collections were created in, and the mother tongue of the users who had added them in their collections. You can explore the diagram by yourself.

This image here shows how, for example, resources created in Finland (the orange node in the network) have ended up in collections of users who speak Hungarian, Estonian, Lithuanian, etc. as their mother tongue.

Note that this graph does not make any assumption of the language in which these resources are in! If I'm right in my guess, most of these resources were in English, not in Finnish..

But anyhow, I find that as a demonstration that these resources can cross borders of some kind. In this case, a Finn has created the resource. It can be just a very little hint available in the design of the resource that it was a Finn, but still some of the underlying pedagogical assumptions or some hints of Finnish curriculum might be embedded in these resources. Nevertheless, or thanks to that, the resources created in Finland seem like a hit (they are in 8 different language groups).

Ok, to me more truthfully, I think this is because LeMill was create in Finland that many of the Finnish resources are shared.

About networks of resources and users

This visualisation is to explore the networks of users that form between resources that are shared in collections. I think this is one of the most interesting visualisations of the dataset, and the one that inspires me the most.

Same as before, click to interact within the image, or if you click on the title on top of the image, you can get the network in a bigger window.


What's there? It's a network diagram where the nodes represent users (user id number) and the edges are the names of learning resources that these users have saved in their collections.


You can zoom into the diagram and explore it. Same as with the previous post, we can see that lots of the resources that users have put in their collections are not shared with other users. These are the singletons that are not part of the common network here.

Then, there are some star like structures that can be found. Like this one. Here the resource highlighted is something that both users (user 59 and 155) had added into their collection.

What I think, I would almost bet on, is that if these users were made aware that they share this resource in their collections, they would be interested in looking at what other resources are in the other person's collection. In this case the user 59 could be interested in looking at the collection of the user 155 has put in her collection.

This basically would be the idea of making underlying social networks visible in a repository to allow social navigation of like-minded users collections. Or, if you wish, a recommender could take advantage of these underlying connections as well. For the recommender, though, the data is very sparse, as can be seen from the visualisation. For that reason, I think we first should explore social navigation possibilities, and then launch for recommenders, when we get more data.

These resources that connect users, or in some cases (hopefully one day) even communities together, are valuable stuff. I have previously referred to this as one way to identify learning resources that cross borders easily. In this case, the two communities could be speaking different languages or be from different countries.

Some suggested that these objects could be also boundary objects. I cannot get my hands on the original article now (frustration of working from home!), so I am referencing some others that reference it:
Star (1989) and Star and Griesemer (1989), on the other hand, are concerned with the distribution of artefacts across communities. Boundary objects are artefacts used by communities: they cross the boundaries between communities and retain their structure, but are interpreted differently by them. The notion of boundary objects was developed by Star (1989) and Star and Griesemer (1989) as a way to explain co-ordination work between communities.
In a larger sense, maybe some of them could be boundary objects. I will need to think about this more..

Anyway, here is another little visulaisation that is actually an overview of the resources that users have saved in their collections. You can visualise it in many ways, you the ordering function on the top.




Star, S. L. 1989. The structure of ill-structured solutions: boundary objects and heterogeneous distributed problem solving. In Distributed Artificial intelligence (Vol. 2), M. Huhns, Ed. Morgan Kaufmann Publishers, San Francisco, CA, 37-54.

Learning resources as part of collections - what about the network?

I'm just exploring a new dataset that I got from LeMill, it contains information about learning resources that users have put in their "collections". Collections is a tool for users to create their own sub-sets of resources and give them a common title, e.g. I find 5 resources on pyramids, I add them to my collection, and I call it "Pyramids for 5th graders", as I am going to use it during my History lesson that I teach with 5th graders.

I think that collections-tool is an excellent tool, also for me as a researcher ;) What I am interested in knowing is whether we could make the links between these collections visible. The link would, of course, be the resources that are shared with collections.

Let's just explore the early visualisation of LOs connecting the collections. Click on "click to interact", and you get the life image. Alternatively, you can click on the title in the image, and you'll have the whole visualisation in a bigger interface. So what's there?





What you first see is a top-level overview of users' collections using a network diagram. It first looks like a grid; the ones on the top left hand corner are small one, they only contain a few resources. The other ones towards the right bottom corner look more clunky and visibly bigger, they include many more resources and are actually overlapped one with another.

You can start zooming in with your mouse. You see that some names will start appearing. Those are the name of the collection and the resources within. With a right click on your mouse, you see a hand appearing. This allows you to move within the visualisation. What you see here is a huge amount of what is called “singletons” in the network jargon. These singletons are collections, but they do not have any connections through shared resources to other collections.

Now, try to locate yourself in the area where that big cluster is, at the bottom right hand corner.

Now, instead of looking at separate little singletons, we are hoovering over a “giant component”. This is clearly the largest group of nodes within this network and some of them seem interconnected. With interconnection I mean that the same resource is in more than one connection.

You can visualise this nicely, if you click on some of the big nodes. It will be highlighted in orange. This way you can see what are the resources related to this collection (the collection name is the node). Interestingly, you'll see some of the resources act as a connection between different collections.

What we can already quickly see is that something called “middle regions” are entirely missing from this network. They represents rather isolated groups that interact amongst themselves. In our case they would be a few resources that are in a few collections by a few users. There do not seem to be any such "isolated stars" in this network of collections. The cool thing about these isolated stars is that over some period of time, they might merge with the giant component. This would happen through a resource that is shared in both the giant component and the smaller entity.

Ok, visualisation is just a visualisation, a snapshot of a moment. More work is needed to properly analyse what is going on, and most importantly, does this have anything to do with how we can make a repository of learning resources a better place?

Well, I of course am on my SNA trip and think that it can help anything and everything, but more about that later..

Reference

Users, LOs, collections and networks forming

This visualisation is to explore the networks of users that form between resources that are shared in collections. I think this is one of the most interesting visualisations of the dataset, and the one that inspires me the most.

Same as before, click to interact within the image, or if you click on the link, you can get the network in a bigger window.

What's there? It's a network diagram where the nodes represent users (user id number) and the edges are the names of learning resources that these users have saved in their collections.



You can zoom into the diagram and explore it. Same as with the previous post, we can see that lots of the resources that users have put in their collections are not shared with other users. These are the singletons that are not part of the common network here.

Then, there are some star like structures that can be found. Like this one. Here the resource highlighted is something that both users (user 59 and 155) had added into their collection.

What I think, I would almost bet on, is that if these users were made aware that they share this resource in their collections, they would be interested in looking at what other resources are in the other person's collection. In this case the user 59 could be interested in looking at the collection of the user 155 has put in her collection.

This basically would be the idea of making underlying social networks visible in a repository to allow social navigation of like-minded users collections. Or, if you wish, a recommender could take advantage of these underlying connections as well.

These resources that connect users, or in some cases (hopefully one day) even communities together. They are valuable stuff. I have previously referred to this as one way to identify learning resources that cross borders easily.

Some suggested that these objects could be also boundary objects. I cannot get my hands on the original article now (frustration of working from home!), so I am referencing some others that reference it:
Star (1989) and Star and Griesemer (1989), on the other hand, are concerned with the distribution of artefacts across communities. Boundary objects are artefacts used by communities: they cross the boundaries between communities and retain their structure, but are interpreted differently by them. The notion of boundary objects was developed by Star (1989) and Star and Griesemer (1989) as a way to explain co-ordination work between communities.
In a larger sense, maybe some of them could be boundary objects. I will need to think about this more..

Anyway, here is another little visulaisation that is actually an overview of the resources that users have saved in their collections. You can visualise it in many ways, you the ordering function on the top.



Star, S. L. 1989. The structure of ill-structured solutions: boundary objects and heterogeneous distributed problem solving. In Distributed Artificial intelligence (Vol. 2), M. Huhns, Ed. Morgan Kaufmann Publishers, San Francisco, CA, 37-54.

From Attention metadata to Participatory metadada

Capturing and taking advantage of users’ actions on the Web has come a long way since business models were first implemented around the idea of clickstream in the ’90 . Instead of having the commercial sites taking advantage of the attention that users pay to different products, in the recent years the tide has turned arguing that interactions with the content (e.g. buying, listening, reading feeds) and users reactions to that content (e.g. ratings, reviews, tags) should be something that the user can control.

AttentionTrust.org, for example, calls this "attention data" and argues that it is a valuable resource that reflects user’s interests, activities and values, thus serves as a proxy for their attention.

AttentionXML (1) is an open specification to capture individual’s clicks to track user’s behaviour and information consumption on the Web. Contextualized Attention Metadata (CAM) schema was build upon it with an extension that allows capturing observations about users activities in any kind of tool, not just a browser or newsreader (Najjar et.al. 2006a,b).

Attention Profiling Markup Language (APML), on the other hand, offers a way for a user to create a personal Attention Profile, which is portable, sharable and captures users’ attention on self-defined services. Moreover, the social aspect of the Web, where users not only interact with resources, but actually participate in communities and create content, has created a need for users to capture these participatory aspects of their attention.

Thus User Labor Markup Language (ULML) that proposes an open data structure to outline the metrics of user participation in social web services. One of the ULML use cases, for example, is around creating metadata (e.g. tagging, voting, commenting etc.) as a way to improve and maintain users’ existence in social web. All these specifications serve the same goal; being openly transparent about one’s interests on the Web in order to make the best use out of them for the user’s own benefit.

I'm currently thinking with my studdy-buddy Nikos Manouselis how we could save such attention profiles from different repositories to have a more holistic picture of what do users do on educational repositories or on federations of them. I think that alone would be a great advance for the research.

Second, it might be that the same user have profiles in different repositories (like I have one in MELT, in LeMill and OERCommons), so this would allow the user to consolidate her interests and resources found in different places, like bookmarks or collections that I have created in these different repositories. It could be nice to have my personal tagcloud based on my attentions in different repositories to allow me to access resources in these different services this way.

Third, there are resources that many of the educational repositories share. Like in MELT, we have most bookmarks on resources from LeMill. It is of interest for LeMill to know that they have fans and users in MELT, so this is the info that can be fed back from MELT to LeMill, and they can boost their stats with this! Not to mention of getting back the participatory information from MELT, e.g. users tags, ratings, etc.

The fourth advantage could be that using this type of profiled information to see what resources from LeMill have been of use to the "extended community" (e.g. outside of LeMill's own user base). This info could help them to boost their reputation in the network of repositories. If we knew that half of the repositories in the federation actually have users who interact with LeMill resources, that would give LeMill a great boost as an interesting repository to play with, a reputable provider of resources (someone pointed out this saying, hey, think of eBay's reputation points for sellers!). I already had toyed with the idea of "travel well" value for each repository in the federation based on the evidence of previous cross-border use of their resources (of course tracked down using something like portable profile).

Of course, finally, such thing could be used for recommendation purposes and to allow users swiftly find resources of interest without noticing that they have to go to a different repository. Like the previous idea of cross-repository tag clouds.


[1] AttentionXML (2004). AttentionXML specifications, Retrieved June 8, 2007, from http://developers.technorati.com/ wiki/attentionxml.
[2] Najjar, J., Wolpers, M., & Duval, E. (2006a), Attention Metadata: Collection and Management. Paper presented at the World Wide Web 2006 Workshop Logging Traces of Web Activity: The Mechanics of Data Collection, May 23, 2006, Edinburgh, UK.
[3] Najjar, J., Wolpers, M., & Duval, E. (2006b). Towards Effective Usage-Based Learning Applications: Track and Learn from User Experience(s). Paper presented at the IEEE International Conference on Advanced Learning Technologies (ICALT 2006), July 5-7, 2006, Kerkrade, The Netherlands.