The cluster members were interested in the data set created so far by ReMashed to explore how to recommend persons in informal environments based on their competences. Therefore, the new competence fields in the personal profile of the learners are quite suitable. Further, we discussed the need to create a widget / portlet interface for ReMashed to integrate it into personal environments like iGoogle and Netvibes but also into Liferay or Moodle.

Following up on this, we talked about the need create an administration layer above the current systems to create different instances of ReMashed for various online communities. From here we stumbled into a 'privacy issue / suitability of information' discussion for different communities.  Just fetching all public items from delicious and present those in all instances of ReMashed might not be wise. Instead of that the hashtag method like in Twitter or Elgg might be a solution to allocate information with a certain tag to a specific community. For instance the hashtag  ’communityBuilding’ indicates that the attached information should go to a communityBuildiung instance of ReMashed.  

How satisfied are you with the ReMashed system?
What I like about ReMashed is the creation of cumulative data that is related to my topics. It is interesting to follow the contribution of the other users in the various web 2.0 channels. I especially like to have recommendations on top of the sea of information as they point me to information I might missed or even remind me on things I find interesting.

I like the way how ReMashed combines different data sources data sources that makes it different to other system. In case it would offer only text based information it would be to overwhelming. The pictures and slides make the system more attractive.  I’m looking forward to additional media integration.

Sounds like you are quite satisfied, any improvements?
Yes, of course. Just taking the picture example, I would suggest using not only specific services like flickr rather than combine different picture services  like flickr and picassa in one picture category.
I observed that there are high and low level picks for the different services. Regarding the pictures it seems that people upload pictures when they attend conferences or other events. In this time you get a whole bundle of new pictures. In normal seasons the picture section relative quite. Regarding this peak times I wish to have more control over the box size for instance through expand or minimizing them.
Regarding the delicious links, I miss comments to the links. Otherwise it is hard for me to judge how I should rate an item. Thus more information and transparency would be great. Further, I would like to add source that are not written by me but might be relevant for the community. It would be great if there would be an option to add resources I regularly reed on the web.

What is your personal information strategy for ReMashed, how do you use the system?
Well, ReMashed is like the online newspaper of the emerged community. I do not have all participants in my Google reader but I like their contributions. My strategy is to look first for new contributions, afterwards I check if I’m still top of the week  and finally I check the rating- based and afterwards the tag-based recommendations.

How do you rate?
Actually I only rate between 3 and 5 stars. I do not use 1 or 2 star ratings instead of that I would prefer a negative rating (a black star).

What is a 5 star rating for you?
I give a 5 star rating to information that is really new for me, something that as an AHA effect attached to it.  There is not a specific topic connect to my 5 star rating. I rate fun stuff also with 5 stars like industry news. I also use 5 star rating if I want to share the item with others thus to make them prominent.

When do you use a 3 star rating than?
For things I find interesting and want to know more about. 3 stars also serve as bookmark for things I want to discover later on.

Would a bipolar rating like thumb up / down efficient for you?
Actually not, because I would prefer 3 positive stars and one thumb down rating for items I don’t like. That would serve my rating behavior.

There are two things regarding the rating:

1. I would like to have an overview of the ratings I already gave to items. Something like a personal view on rated items.
2. I would like to see an average rating of the community or the amount of users that rated an item.

How does the tag- and rating-based algorithm serve you in the beginning compared to the current situation. Did you recognize any differences?
From the very beginning the tag-based recommendations were really efficient for me. They immediately showed sources of other person that are related to my interests. At the moment they become a bit less relevant. Related to that the rating-based recommendations were not so related in the beginning, but now they seem to work better and actually I like them now more than the tag -based recommendations. Therefore, the tag based recommendations could get maybe a new functionality.

Regarding a new functionality we currently consider to use specific learning goals to recommend items that support users in their competence development. We described in that in the last posting [you can find it here]. Can you imagine to rate items regarding your competence development goals?
Well, for me such a rating mechanism has to be as informal as possible becasue I use the system for pleasure. Thus, it should be free of policies or specific behavior model I have to follow to get pedagogic sound recommendations.  I could imagine defining specific goals with keywords or a tag cloud. This specification could influence the tag-based recommendation approach. It could recommend items that serve this learning goal. Through the rating of the items they can be weighted and become more or les important in the tag-based recommendation approach.

Like a  combination of tag- and rating-based recommendation.
Yes something like that.

Many thanks for your time and for the interview.
It was a pleasure for me.

• 40 users are registered (+ 25%)
• 4261 items are available in the system (+ 26%)
• 220 items are rated (+ 71%) 
• 713 recommendations are offered (+ 38%)

Those are great numbers ReMashers!

Latest News

Luckily David Wiley visited us at OUNL, Marco and I had the opportunity to talk to him about the Open Content issues. We also presented to him the ideas behind ReMashed and further development plans. David was pretty excited about the system and compared it with their OCW finder. We were happy to find him a day after our meeting as a member of ReMashed.

Other observations

The rating based algorithm seems to come to a threshold that makes its recommendation more reasonable. ReMashed user have now the opportunity to experience how the algorithm works in small communities. For example, I get now recommendations for items I never rated but my neighbors did. As five ReMashed users sitting in the famous ApeCage at OUNL we share our experiences with the system with each other. I complained about one rating-based recommendation I received and was wondering who could have rated such an item. My colleague Marco told me that he rated it just for fun and now I have to live with this recommendation. Arrgh! (Reminds me to add the rating of recommended items functionality to the system).

More room for improvement

I got various requests how people can rate an item that is recommended to them when they like it. As I said we are working on that update but it will come with different other updates. For now you have to follow the recommended item to its original source, tag it in delicious and rate it afterwards when it appears in ReMashed (Uhhh bad user interaction). In a future future release we have to create the possibility to add sources from ReMashed directly to delicious we added it to the ToDo list.

Export recommendations via RSS

Another request is the export of the recommendations via RSS. Three people already asked if they can subscribe via RSS to the recommendations of ReMashed. That’s a nice and important idea that we will keep in mind for future development and also added it to the ToDo list.

Future development

Talking about future development, I can present an initial mock-up of the next release were users can specify additional Web2.0 sources (twitter & Youtube) and define interests fields (learning goals) with a self-assessment slider. The interests fields will be taken into account for future recommendation regarding personal competence development. The idea is that a user can explicitly specify 1 to 3 interest fields (learning goals) and his related competence level between 1-5.  An additional recommendation algorithm will be triggered by this goals and present relevant items for future competence development to the user in a separated box.

As a smart edition the interest fields will be powered by an auto completion algorithm. This algorithm will be fed with learning goals of other users and tags that are available in the system. Using auto completion helps to support users to use a shared vocabulary regarding their competence development. Later on users can be grouped according to their shared learning goals. More tailored recommendations can be created through recommending resources of users that are on a higher competence level than the current user.

Using domain data sets to cover cold-start and add additional techniques

Through the current pilot we came across a new idea to create different domain data sets with rated items to cover the cold-start problem of the recommender. The running pilot is a good example of a Technology Enhanced Learning data set. If we could run different pilots in various domains we could create several domain data sets (medical, engineering, education, security etc.)  and apply them to cover the cold-start and add additional technologies like Latent Semantic Analysis.

If anyone is interesting in cooperating with us in additional ReMashed pilots don’t hesitate to contact us.

Some descriptive statistics:

• 32 users are registered from Japan, USA, Netherlands, Estonia, Canada, Austria, Germany and Sapin (people still start joining )
• 3363 items are available in ReMashed
• 128 items are rated so far (only! a real cold-start)
• 501 recommendations are offered.

First experiences:

A nice effect of ReMashed is the recommendation of 'older' information that is realted to my current information needs. Whenever new people sign up to the system my personal profile in ReMashed gets sources recommended that they created in the past but that fit to my current information needs.
For my own evaluation I use a developer, my own and a educational account to assess the recommendation algorithms. Each of the accounts receives tailored recommendation from the system regarding their interests. Thus, the personalization of the sources works fine. For the next release I want to integrate competence level of users and offer tailored recommendations with a context-aware recommendation algorithm.

Also interesting is the frequency of use of particular services. Most updates happen over delicious and blogs. Slideshare and flickr are used less. I’m planning to weight blog posting and slideshare contributions to make them more important for the recommendation technology.  Blog posting and slides from slideshare are less frequently used but they offer a higher quality contributions as people put a lot of effort into them. We still believe that people blog about topics they are interested in whereas they also tag information on delicious which are not part of their competence development profile. In order to create a recommender system (service) for informal learning environments these sources have to be weighted stronger to prevent that they are neglected by the overwhelming contributions of delicious bookmarks.

Many thanks to the community around ReMashed! Please keep the spirit and forward any idea or comment regarding ReMashed to me. You really help to improve the system and the underlying recommendation technologies.

The most frequent questions are:

Should I rate my own contribution?
Sounds strange but yes, please rate your own contribution to explain to the rating based algorithm that these are the topics you are interested in. At the moment ReMashed still suffers the cold-start problem of recommender systems. Thus, there are too few ratings in the system to make sufficient rating based recommendations. However, this is why we decided to go for a hybrid recommendation strategy and use the tags of the contributions as a second  recommendation technique. This 2nd technique (the tag based recommendation approach) works quite adequate. It looks like the tag based recommendation approach does a good job in covering the cold-start of the rating based algorithm. Maybe I should implement a threshold of ratings before I enable the rating based algorithm.

For the user of ReMashed please rate as much as possible! Also when you enjoy a information only a bit give it 1 or 2 star rating to feed the rating based algorithm to offer more relevant recommendations.

I miss the possibility to rate the recommendations I get.
Yeah, me culpa you are absolutely right. I also want to have more control over the recommendations that are offered to me. In the next release there will be the possibility to rate the recommendations.

The first week gave us already a lot of ideas for the future.
Have fun using ReMashed.

So in short, we had a great week with lots of new learning experiences. Most of the sessions were longer than expected because the participants continued the discussion of the presented topics.

On Monday, I gave together with my colleague Christan Glahn a session about Mash-up environments. The objective of this session was to identify potentials of the different mash-up techniques for personalized learning environments (PLEs). The session was organized in two parts. We discuss how mash-up systems are related to learner support, context-awareness and self-directed learning and recommendation strategies; and presented the  prototype mash-ups (ReScope and the ReMashed system) which focus on the support of learners in informal and self-directed learning.

A rather interesting experience was the late night session on PowerPoint karaoke. PP Karaoke is a very good training for giving presentations, because you have to stick to rules of how to present efficiently and you learn how to improvise during the session. Skills you also need for serious presentations.You can find a picture here.

There are no mandatory rules for PowerPoint Karaoke. To get the ultimate experience some principles should be respected:
•    The presenter shall NOT see the presentation slides in advance.
•    The presentation time shall be limited. We used 10 minutes as a time limit.
•    The jury shall have a sense of humor.
•    The speech must be related to the Powerpoint slides. General nonsense is not allowed.
Try it out on your next event; there is also an Open Source tool that uses the Slideshare API. http://www.slidesharetoys.com/karaoke/

The strangest experiences at the TCWS09 were the Monsters from the Monster Academy. I was quite surprised how well informed they are about research in TEL. One of them Roddy Rube ask me to give an interview about my presentation. I hope it will be viewable soon on the Monster Academy channel. For now you get a picture of the interview. 

Now it’s time for bug fixing.

Please find here the recording of the questions-answer round at the RecSys08 at the end of the Doctoral Symposium. Joseph Konstan and Alexander Felfernig discuss questions of Phds towards recommender systems. Interesting questions regarding decision support, psychology effects around decisions, and a pragmatic view on recommender systems are discussed. Feel free to download the files Alexander and Joseph agreed to publish them.
You can decide between a 24 MB wma file and a 50 MB mp3 file.

Have fun with this informal learning activity on Recommender Systems.

On Sunday the last day of the RecSys08 was the Doctorial Symposium. The academic world was represented by Pearl Pu (Switzerland), Joeseph Konstan (USA), Alexander Felfernig (Austria) and Francesco Ricci (Italy). 9 PhD projects were invited to present their recent results and to get feedback for their further development. The invited PhDs beside me were: Olga Santos (Spain), Danielle Lee (USA), Erich Teppan (Austria), Marcos Domingues (Portugal), Akhmed Umyarov (USA), Linas  Baltrunas (Italy), Leobino Sampaio (Brazil) and Youngok Kwon (USA). It was a very good an intensive day with lots of discussions which I missed a bit during the main conference.
The follwing slides I presented in the Symposium:

The overall comments on my project were quite positive. The Symposium enjoyed the clear structure and the experimental design of the studies in my PhD project.
Alexander suggested me to apply also model based recommendations like Bayesian predictions for my recommendation task. Actually, in the beginning of my project we decided to focus on memory based techniques and for collaborative filtering. However, after having some experiences with recommenders in TEL I agree with Alexander that model based predictors are also suitable. Alexander further asked for the development of new algorithms and I explain that the focus of my research is on using existing algorithm in the TEL field to finally find the most suitable for informal learners.
Joseph stressed the aspect of recommending learning items and sequence of items (based on the emerging behavior of the learners in a Learning Network - a learning path). He emphasized the challenges of recommending sequence of items and further explained that he could imagine that the TEL field might offer some interesting findings because differently to product recommender the world of education is quite well known. Finally, he stressed the informal learning aspect of my project. He highlighted that in informal learning you never know the structure of a learning item like in formal learning offers. He further described differences of the recommendation task in informal and formal learning like I did in one of my papers . The big challenge therefore is to mine the purpose of an informal learning item. Joe could imagine to analyze the purpose of a particular learning item based on sequences of learning items. A very inspiring idea, I haven’t thought of so far.

The DS was really great and I only can encourage PhDs that are working on recommender systems to present their ideas in this setup.

Joe and Alexander with closing comments