I always get a jolt of information visiting ISI. Here are five pointers to things I learned this time:

1. The Karma system [github] is really leading the way on bringing data integration techniques to linked data. I’ll definitely be looking at Karma with respect to our development of the Open PHACTS platform.

2. I’m excited about change detection algorithms in particular edit distance related measures for figuring out how to generate rich provenance information in the Data2Semantics project. These are pretty well studied algorithms but I think we should be able to apply them differently. A good place to start is the paper:

• Sudarshan S. Chawathe, Anand Rajaraman, Hector Garcia-Molina, and Jennifer Widom. 1996. Change detection in hierarchically structured information. SIGMOD Rec. 25, 2 (June 1996), 493-504. DOI=10.1145/235968.233366 http://doi.acm.org/10.1145/235968.233366

3. Also with respect to provenance, after talking with Greg Ver Steeg, I think Granger Causality and some of the other associated statistical models are worth looking at. Some pointers:

• Information Transfer and Social Media by Greg Ver Steeg, Aram Galstyan
• Granger causality and transfer entropy are equivalent for Gaussian variables by Lionel Barnett, Adam B Barrett, Anil K. Seth

4. Tran Thanh gave a nice overview of his work on Semantic Search. I liked how he combined and extended the information retrieval and database communities work using Semantic Web techniques. Keyword: Steiner Trees

5. MadSciNetwork is a site where scientists answer questions from the public. This has been around since 1995. They have collected over 40,000 answered science questions. This corpus of questions is available at MadSci Network Research. Very cool.

Finally… it’s nice to visit southern california in January when you live in cold Amsterdam

The class involves 6 lectures covering the major ground of Semantic Web technologies and KR. We then give them 3 1/2 weeks to design and hopefully build a Semantic Web application in pairs. During this time we give one-on-one support through appointments. For most students, this is the first time they’ve come into contact with Semantic Web technologies.

This year they built applications based on The Times Higher Education 2011 World University rankings. They converted databases to RDF, developed their own ontologies, integrated data from the linked data cloud and visualized data using sparql. I was impressed with all the work they did and I wanted to share some of their projects. Here are four screencasts from the applications the students built.

Points of Interest Around Universities

Guess Which University

Find Universities by Location

SPARQL Query Builder for University Info

At the VU, we are also working with Elsevier Labs on the Data2Semantics project where we are trying to enrich data with additional machine understandable metadata. How does this relate to metrics? I believe that metrics (access, usage, etc) can be e a key piece of additional semantics for datasets. I’m keen to see how metrics can make our data more useful, findable and understandable.

The issue highlights top research at the intersection of provenance and the Semantic Web. The papers addressed a range of topics including:

• tracking provenance of DBpedia back to the underlying Wikipedia edits [Orlandi & Passant];
• how to enable reproducibility using Semantic techniques [Moreau];
• how to use provenance to effectively reason over large amounts (1 billion triples) of messy data [Bonatti et al.]; and
• how to begin to capture semantically the intent of scientists [Pignotti et al.].

A common thread through these papers is the use of already existing provenance ontologies. As the community comes to an increasing agreement on the commonalities of provenance representations through efforts such as the W3C Provenance Working Group, this will further enable new research on the use of provenance. This continues the fruitful interaction between standardization and research that is one of the hallmarks of the Semantic Web.

Overall, this set of papers demonstrates the latest approaches to enabling a Web that provides rich descriptions of how, when, where and why Web resources are produced and shows the sorts of reasoning and applications that these provenance descriptions make possible

Finally, it’s important to note that this issue wouldn’t have been possible without the quick and competent reviews done by the anonymous reviewers. This is my public thank you to them.

I hope you take a chance to take a look at this interesting work.

The tutorial I gave focused on exposing relational databases to the web using the awesome D2R Server. It’s really a great piece of software that shows results right away. Perfect for a tutorial. I also covered how to get going with LarKC and where that software fit in the whole semantic web data management space.

On to the story…

The students easily exposed our test database (GPCR receptors) as RDF using D2R. Now the cool part: I found out just before the start of my tutorial  that the day before they had setup an RDF repository (Sesame) with some personal profile information. So on the fly I had them take the RDF produced by the database conversion and load that into the repository from the day before . This took a couple of clicks. They were then able to query over both their personal information and this new GPCR dataset. With not much work we hand munged together two really different data sets.

This is old hat to any Semantic Web person, but it was a great reminder of how the flexibility of RDF makes it easy to mashup data. No messing with about with tables or figuring out if the schema is right, just load it up into your triple store and start playing.

Thanks to Shiyong Lu for the pointer.

The program that Graham has created with Y-combinator, in my opinion, resembles in parts a PhD program. Like a good PhD program, Graham focuses on identifying and then developing people. The goal is to create a founder that can develop a great business. Similarly, the goal of a PhD program is to develop researchers who can do great research. In both cases, the idea is to provide the tools, support and environments that make great business founders or great researchers. I think in both cases we are the business of

people development – not product development.

However, I think Y-combinator offers more than just people development [1], it creates a creative, intense, social atmosphere. It’s designed to not only create talent but teams of talent. There are not many places in the world that talented people just work on the project their interested in – together.

Graham compares the effect to a coral reef, a self-generating ecosystem whose members provide nourishment for one another. “Pick the right founders and help them—and the coral reef will just happen,” he says

This brings me to the next article about the Kinnect. There were several ancedotes within the article about how Phd students quickly embraced the Kinect to do interesting and novel stuff: Here’s an example:

A group from UC Berkeley strapped a Kinect to a quadrotor—a small helicopter with four propellers—enabling it to fly autonomously around a room. A couple of students at the University of Bundeswehr Munich attached a Kinect to a robotic car and sent it through an obstacle course.

You need two things to pull of something like that:

1. free time
2. talent

PhD programs have a tendency to have these properties. Talented people with enough spare bandwidth to try new things (although I have a feeling phd students are becoming too busy lately…). As you noticed from the quote, it was a team of students that did the work. From my own personal, experience this happens all the time in phd programs. Indeed, there are great examples such as Google, Yahoo and Dropbox. But I think in general we in academia see this as happy side effect and don’t encourage or even look at it is a goal of our PhD programs.

Instead, I believe that we should actively encourage such projects. PhD programs could become talent incubators. We in the academy can encourage the generation of talented teams, for example, making group projects a key part of phd programs. Indeed, one could see a future where universities specifically advertise or promote project teams (groups of students) to business or investors as success-ready talent. Additionally, this would also be an interesting way of providing alternative careers to PhD students who want to do something outside activities.

Thoughts?

[1] Clearly *people development* is the one of the key services that universities provide. In general, i think universities are good at people development (although we could always be better) However, as the demand for high education increases and the pressure on universities to be more cost effective, I think it’s pertinent to begin to think about where universities can have added value that other higher education delivery mechanisms can’t. This is what this post is about.

The aim of the workshp was to gather together the group of people working on next generation measures of science based on the Web. Importantly, as organizers, Jason, Dario and I wanted to encourage the growth of the scientific side of altmetrics.

The workshop turned out to be way better than I expected. We had roughly 36 attendees, which was way beyond our expectations. You can see some of the attendees here:

There was nice representation from my institution (VU University Amsterdam) including talks by my collaborators Peter van den Besselaar and Julie Birkholtz. But we had attendees from Israel, the UK, the US and all over Europe. People were generally excited about the event and the discussions went well (although the room was really warm). I think we all had a good time the restaurant, the Alt-Coblenz – highly recommended by the way-and an appropriate name. Thanks to the WebSci organizing team for putting this together.

We had a nice mix of social scientists and computer scientists (~16 & 20 respectively). Importantly, we had representation from the bibliometrics community, social studies of science, and computer science.

Importantly, for an emerging community, there was a real honesty about the research. Good results were shown but importantly almost every author discussed where the gaps were in their own research.

Two discussions come to the fore for me. One was on how we evaluate altmetrics.  Mike Thelwall who gave the keynote (great job by the way) suggests using correlations to the journal impact factor to help demonstrate that there is something scientifically valid that your measuring. What you want is not perfect correlation but correlation with a gap and that gap is what your new alternative metric is then measuring. There was also the notion from Peter van den Besselaar is that we should look more closely our how our metrics match what scientists do in practice (i.e. qualitative studies). For example, do our metrics correlate with promotions or hiring. The second discussion was around where to go next with altmetrics. In particular, there was a discussion on how to position altmetrics in the research field and really it seemed to position itself within and across the fields of science studies (i.e scientometrics, webometrics,virtual ethnograpy ). Importantly, it was felt that we needed a good common corpus of information in order to comparative studies of metrics. Altmetrics has the problem of data acquisition. While some people are interested in that others want to focus on metric generation and evaluation. A corpus of traces of science online was felt to be a good way to interconnect both data acquisition and metric generation and allow for such comparative studies. But how to build the corpus….Suggestions welcome.

The attendees wanted to have an altmetrics12 so I’m pretty sure we will do that. Additionally, we will have some exciting news soon about a journal special issue on altmetrics.

Some more links:

Abstracts of all talks

Community Notes

Also, could someone leave a link to the twitter archive in the comments? That would be great.