Friday, April 15, 2005

content-free research

This story (link from The Corner) is funny:
Jeremy Stribling said Thursday that he and two fellow MIT graduate students questioned the standards of some academic conferences, so they wrote a computer program to generate research papers complete with "context-free grammar," charts and diagrams.
The trio submitted two of the randomly assembled papers to the World Multi-Conference on Systemics, Cybernetics and Informatics (WMSCI), scheduled to be held July 10-13 in Orlando, Florida.
I can't say I'm astounded. Those jargon-loaded treatises with ten-mile sentences are a bit opaque. I've often had the experience of reading half-way through the first page and realized that I had no idea what they were talking about. To me, this means that the paper is poorly written, but there are lots of people in academia who take this as a sign that the authors are just too smart to understand. A referee with this attitude might decide to recommend the paper just to show how smart he is in being able to recommend such smart papers.

By the way, the AP reporter was confused about the "context-free grammar". That isn't something that goes into the paper like charts and diagrams, it's the technique that was used to generate the paper. This is pretty impressive because context-free grammars are very weak generators. I would expect a project like this to require a context-sensitive grammar.

(Yes, I know I wrote "content-free" in the title instead of "context-free". It's a computer science pun. We computer scientists are a crackup.)

UPDATE: On further thought, it occurs to me that maybe they did mean that they put context-free grammars in the papers. Such things are common in computer science. Probably the primary reason I thought "context-free grammar" described the generator rather than the contents of the paper was because it was singular, whereas "charts" and "diagrams" were both plural. Also, the qualifier "context-free" is a bit odd there, and why not other code? I'd expect something more like "grammars, code fragments, charts and diagrams". On the other hand if you wanted to impress insiders then you would tell them that you generated the papers with a grammar that was only context-free.

If I have time this weekend, I'll try to find out...

No comments: