On Wednesday, there were tutorials (Semantic mashups with Wandora, Improving Microsoft Sharepoint with Topic Maps, Hands-on TMQL, and Hands-on TMCL with Onotoa) and a one-day code camp for people interested in working with the Ontopia source code.

Thirty or so people attended the code camp. Lars-Marius Garshol and Geir-Ove Grønmo, who ran it, had
looked at the list of registrants in enough detail to realize, in the week before the meeting, that they would not be able to assume that everyone attending had a good grasp of Ontopia and its parts even from a user perspective, still less from the perspective of a coder and bug-fixer. So they began the day with a general overview of the system and its parts, which was quite useful. Towards the end of the morning, they canvassed the attendees for particular topics people wanted to discuss, and after lunch they addressed those topics using slides they had gathered or created during lunch. The second half of the afternoon was devoted to individual work. Possibly some participants spent time trying to diagnose one of the bugs marked ‘suitable for newbies’ in the Google Code issues list; me, I spent it finally getting Ontopia’s web interface to run on my machine. (It turns out that there IS documentation on how to do that and how to get started with Ontopia after you install it — there just wasn’t any documentation that told me where to look for that getting-started document. When you check out the Ontopia source, there are instructions on how to use the source to compile a fresh build. Once you do that, look INSIDE THE BUILD YOU JUST MADE for documentation, NOT in the ./doc directory higher up in the source tree. The document you want is build/dists/ontopia-5.0.2/doc/install.html [as long as the current build is 5.0.2].) The build has a copy of Tomcat all set up for use; all you have to do is launch it, and point your browser at it, and you can use the various Web-based tools running locally.)

Wednesday night, Lars-Marius and Geir-Ove and some others of us went to a brew-pub where they wanted to check out the seasonal beers. (Both are serious beer tasters, and Geir-Ove also brews as a hobby.)

On Thursday, we had first the opening keynote, in which I talked mostly about the (to me, initially unexpected) relevance of the classic AI hill-climbing problem to the challenge of encouraging wide adoption of a technology. The gist of my argument is that if we regard the choice of computer technology as an attempt to optimize the utility, or efficiency, or whatever property of our hardware and software, then the choice of information technology is an example of a hill-climbing problem. People change their choices in an attempt to make things better for themselves. The challenge, for those interested in making Topic Maps or any other technology universal, is (a) that most users use the very simplest, most primitive algorithm known for the hill-climbing problem: choose a direction at random, and see if moving a small distance in that direction moves you uphill; if so, do it and try again. Never ever go downhill, even a little bit. This works fine when the surface of the optimization function is smooth and has a single maximum, but it will get caught in a local maximum whenever the surface is not smooth. The refusal to go downhill means, essentially, that no one wants to put up with any temporary loss of functionality in the hopes that things will get better later. This means that any technology you want to sell had better start showing a positive return on investment very fast — which means you can’t require the user to learn very much before they start seeing payback. Also (b) even a positive return on investment is not necessarily enough, since many users regard every change in their computing environment as very painful. There has to be a LOT of good returns on the effort of making any change, to make users feel they have come out ahead. Among the consequences of this state of affairs are that even modest difficulties for new users can deter adoption of a new technology. (To cite just one example: I’m a geek, and I’m fairly self-reliant. But after I downloaded Ontopia’s source code and compiled it, when I couldn’t figure out what to do next, my work with topic maps and Ontopia came to a complete stand-still for about eight weeks.) It’s not always easy to find a path for new adopters of a technology in which every new investment of time, money, or effort brings new returns; actually, it’s very very hard. I think that’s one reason so few technologies ever become ubiquitous. All in all, it’s enough to make me, for one, tend to distrust plans that require universal adoption of a technology in order to work. In the case of XML, for example, universal uptake was not an essential goal: those involved wanted something they could use themselves, and the project would have counted as a success even if XML had never had the wide adoption it got.

For most of the rest of the day, the conference was tracked. I heard Lars Johnsen talk about his wish that Topic Maps might play a role in national data standardization in Denmark (he believes that some problems in the current state of things would be avoided if topic maps were used). Motomu Naito of Kyoto University then talked about his work on constructing authority files for names of members of the Japanese nobility in the Meiji era, using topic maps. The problems are severe, and topic maps seem to provide a good foundation for solving them. Quintin Siebers of Morpheus Kenntnistechnologie BV talked about work they are doing in a project called idSpace, for distributed collaborative work in product innovation. I got dizzy from some of the buzzwords, and his demo turned out to require more bandwidth than the meeting place was in a position to provide (memo to meeting organizers of the world: when you have seventy-five information technologists in a building, all wanting to use the network pretty much non-stop, sharing a single DSL line among them really is not going to do the job). Once he had finished, he sat next to me and reworked the demo, so that during the break he was able to show those lingering in the room that a text-only version of the demo worked. The morning ended with Shu Matsuura of Tokyo Gakugei University talking about a system he is building using topic maps for cross-disciplinary e-learning.

After lunch, I attended a session on Query and Update. Rani Pinchuk of Space Applications Services presented work he and colleagues have done in a question-answering system using topic maps, specifically their work on determining the focus of a question. What is the user asking? What kind of information will count as an answer? This can be hard even for humans; it’s not surprising that machines sometimes have trouble. Lars-Marius Garshol of Bouvet ASA then described some ideas for using topic maps to improve searching, which he has experimented with in an interface to his online collection of photos. I believe I’ve heard him talk about this before, at Balisage, but I think the ideas are maturing. Oddly enough, many systems for searching topic maps seem to use full-text search on the topic map to find things to show the user; they don’t always even try to exploit the structural information in the topic map. As I understand it, the approach Lars-Marius is experimenting with first looks to see whether the search terms provided by the user actually match any names of topics in the topic map; if they do, then the system figures out how the things named might be connected. Given a search for “photos of sam oh”, for example, the system recognizes “photos” as a topic type (in the sample application) and “sam oh” as the name of a person. Searching for things related to both of these topics leads to photos (instances of the topic type) depicting Sam Oh (an ISO WG member). Presumably in a photo collection with many photographers, it might also find photos taken by Sam Oh. Ideally, the interface could choose one interpretation or the other, and let the user confirm the choice or say “no, I meant the other interpretation”. Kal Ahmed of Networked Planet talked about TMSPARQL, and Lars-Marius talked about recent work adding an update facility to the tolog language. (Tolog is scheduled to go away in favor of TMQL, eventually, but TMQL is still in flux, and in the meantime tolog is there, he has an implementation, it works, and he needed updates.)

My memories of the rest of the afternoon are a bit hazier; there were some interesting talks about applications, and a nice session of impromptu talks under the rubric “Open Space session”. The talks were interesting, but I didn’t take notes and I can’t now remember who talked about what.

Friday I mostly remember the poster session, with posters about TMQL4J (a Java-based TMQL engine, which is trying to track the revisions of the draft spec), TM-based e-learning environments, the use of CouchDB and Javascript with topic maps, and one on Implementing TMQL for Teaching. The poster session ended early, so we all sneaked back into the other track to hear Graham Moore of Networked Planet talk about their Web3 platform. As always, he gave a dynamite presentation and persuaded me that they have their act together and know what they are doing, know what market they are trying to serve and what it needs. But the market they are aiming at is apparently far enough away from me that I have trouble remembering any details.

After lunch, another open-space session, and a presentation of the entries in a Best Topic Map of 2009 contest sponsored by Space Applications. It was also many people’s first glimpse of , an online topic-map browser recently launched by the Topic Maps Lab in Leipzig. The audience voted Shu Matsuura’s e-learning topic map this year’s winner. I found the contest and the quick tour of the topic maps surprisingly fun and found myself thinking about what topic map I might be able to build in time to enter it in next year’s contest. (Getting people to think along those lines was, I assume, part of the point.)

Steve Newcomb gave a dynamite closing keynote, with a memorable description of buying a circus tent with a ten-meter center pole, and discovering that it came with no instructions (he infers that most people who buy circus tents already know how to put them up). Putting up the circus tent turns out to involve many people pulling on ropes in different directions - sounds a bit like a tug-of-war - and anyone who has spent much time in standards committees will see a natural analogy with standards work. With luck, people pulling not just in different directions but in opposite directions can manage to get a tent to go up. And looking around the conference, with its mix of academic and commercial applications, and its variety of papers, Steve felt able to say “The tent is up, and the Topic Map circus has come to town.”

Friday night a small group of us went to hear the motets in the Thomaskirche at 6 o’clock (the choir of the Thomasschule is deservedly famous), followed by a concert in the Gewandhaus (Christopher Hogwood conducting Mozart and Richard Strauss). Both were very beautiful, and made a very satisfying end to the week.

Anyone interested in topic maps will find TMRA well worth their while.