Ontology engineering is hard

You don't hear this very often. That's because everyone who's ever tried to do anything more complex than a tutorial has killed themselves.

At present, I'm particularly annoyed by:

• rdf:List and rdf:Seq. If I want to describe a closed container, I have to use rdf:List. However, this makes my ontology into OWL Full, which is undecidable. Otherwise, I have to break the open-world assumption by using rdf:Seq. Author lists are closed containers. What am I to do?
• Reification. I understand reification, and I use it. However, it increases the cognitive complexity tremendously. In a bibliography, for example, papers are published in proceedings or journals. However, it is the proceedings volume that is published (not the paper), and the paper itself does not have a page range or a publication date. An abstract entity (a “publication”) is possessed of a date, publisher, etc., and is associated with what is actually printed (the proceedings). Some other entity associates itself between the paper and the proceedings, linking the two with the page number. This can get awkward: a paper in proceedings now has
  1. a paper, with title etc.
  2. an “inproceedings” reification
  3. a proceedings volume, linked to the conference
  4. a “publication” reification for the printed volume
  5. a publisher, attached to the publication reification
  6. a publication date, also attached to the publication
  7. a page range, attached to the “inproceedings”
  Full details, suitable for a BibTeX record, have to be harvested from the graph. Furthermore, I cannot describe in an ontology that “a tech report is a bibliographic item where the publication reification has a publisher property which is an institution”.

No wonder people just do straight translations of BibTeX, rather than extracting the underlying semantics.

It's not helped by Protégé crashing, not letting me relate to other ontologies (such as FOAF), and spazzing up the loading of files. Back to Vim.

Posted at 2004-06-22 09:58:28 by Richard • Link to Ontology engineeri…
Comments, trackbacks.

Nice phrases

More joy from my referrers logs: I'm the number 1 Google result, both in the US and Brazil, for “nice phrases”, for this short post. Amazing.

Posted at 2004-06-22 08:00:54 by Richard • Link to Nice phrases
Comments, trackbacks.

DEVONthink tips

It's amazing what you pick up going through referrer logs. Here's an interesting way of using DEVONthink, from Matt Henderson:

Of particular importance are the root-level DailyNotes/ folder and the project-level History/ folder.

At all times, I keep a separate window of the DailyNotes/ open, and set to the "Outline" view. This splits the window into two vertically stacked sections: 1) the top section listing the contents of the folder (DailyNotes), and 2) the bottom section listing the contents of any selected item in the above list.

Whenever anything noteworthy happens during the day, I create a new text or RTF document in DailyNotes, and give the document a name that completely describes the event.

Go and have a look, it's a good idea.

Posted at 2004-06-22 07:57:18 by Richard • Link to DEVONthink tips
Comments, trackbacks.

Evil, sodding printers

Our G3 server is connected to a Samsung ML-1451N printer. Quite apart from its total inability to do the claimed network printing for Macs that were made after 1998, it also seems unable to get margins right. Print any file: the top margin is huge, and the bottom margin is tiny. This behaviour occurs for PDFs, PSs, and printing from apps or lp. It happens over the printer share, or locally. It happens regardless of paper tray. It doesn't happen for the demo printout (holding a button on the printer). Thus, I reckon it's Samsung's fucking shit drivers. And now they've broken the download link for their PPD file.

Rule of thumb: never rely on Samsung to get anything right. My ML-1210 at home works fine, but the one attached to our Linux machine starts crying when printing from OpenOffice.org. You have to export as a PDF and print through Konqueror. How crap is that?

Does anyone have any ideas?

Posted at 2004-06-22 03:45:40 by Richard • Link to Evil, sodding prin…
Comments, trackbacks.

Cars and trains

From Peeve Farm comes this quote from the Telegraph:

Encouraging travellers to switch from cars and airlines to inter-city trains brings no benefits for the environment, new research has concluded.

Challenging assumptions about railways' green superiority, the study finds that the weight and fuel requirements of trains have increased to the point where rail could become the least energy-efficient form of transport.

Engineers at Lancaster University said trains had failed to keep up with the motor and aviation industries in reducing fuel needs.

They calculate that expresses between London and Edinburgh consume slightly more fuel per seat (the equivalent of 11.5 litres) than a modern diesel-powered car making the same journey.

The car's superiority rises dramatically when compared with trains travelling at up to 215mph.

Brian Tiemann adds:

There's still the question of traffic congestion if everyone drives, and rail is still cheaper. But rail is also way slower, way less flexible, and (at least in places like, say, San Jose) you still have to drive to the station ten miles away, park, ride, work, ride back, get in your car, and drive home. Which I daresay would add a fair amount to the equation, yet more in favor of four wheels.

I don't think it's true that rail is cheaper. Bear in mind that Top Gear recently went out and bought 3 used cars and drove them to and from Manchester. Including the purchase of a car it cost less than £100 each*, while a rail ticket is £150. They also drove a diesel Audi to (and, I believe, back from) Edinburgh from London on one tank.

Of course, they do say per seat in the article, but consider how empty most buses and trains are outside of peak times.

It's certainly cheaper for me to drive to Reading each day (just under £5 in my luxury saloon, listening to my own music, to my door) than it would be to take the train (around £7 for the train, plus £2 on the buses, plus walking, plus waiting). If I switched to something like a diesel Honda Civic, or a Civic Hybrid (thus doubling my fuel-efficiency), I'd be spending vastly less to have a faster, more comfortable journey.

Car-sharing, it costs the same £5 to take 2 people, rather than £18 on public transport. There is no competition at all.

Why on earth would I take the train? Why would anybody?

* Much, much less, I think, but I don't want to be wrong. I recall something like £50, and one of the cars only cost £1.

Posted at 2004-06-22 02:53:33 by Richard • Link to Cars and trains
Comments, trackbacks.

XHTML Google Search Box

For the benefit of the GoogleBot: validate XHTML Google AdSearch WebSearch search box HTML capital table. There, I feel much better.

I received an email this morning from Google. I'd sent them an email saying, essentially, my page is XHTML1.1; your generated search code comes from 1996, and even then it wouldn't validate. They've said that I am permitted to alter the generated WebSearch code in order to make the code XHTML compliant, which I think is pretty nice of them.

So, I've renamed all of the capitalised elements, quoted all the parameters, deleted the parameters that haven't existed for years, deleted all the table tags (they were totally unnecessary), etc. It does make me wonder who on earth generated their code; even Google.com isn't that bad (though they don't quote their attribute values).

I have to admit, I did also shrink the form entry area, as it was far too wide to go in my sidebar.

Thoughts? I actually like it a lot; I get all my search options at the top, a tiny Google logo (because I haven't set one), and almost no colour on the page. OK, there are ads, but no sidebar!

Posted at 2004-06-22 01:41:37 by Richard • Link to XHTML Google Searc…
Comments, trackbacks.