Tuesday, 26 August 2008

Context-sensitive constructions in English

Written by Martin Kleppmann on Tuesday, 26 August 2008, 13:25 GMT.
Filed under: techie notes.

Anybody who has read about basics of computational linguistics (i.e. the cross-over area between language and computing) will have come across the (very useful) definition of a context-free grammar. Many computer-readable languages such as programming languages are designed to be context-free, since this makes them more manageable for both the computers and the humans operating them.

There are also grammars which are more powerful than context-free grammars, and which allow more complex statements; the next level of complexity are context-sensitive grammars.

Ever since these mathematical definitions were made, people have wondered how our human languages fit into this framework. There has been a lot of discussion about this in the academic community, and currently the general opinion seems to be that most human languages are more or less context-free, but there are a few exceptions, such as a context-sensitive construction in Swiss German.

In my work at the moment I am trying to help improve the accuracy of computer understanding of certain statements in natural language. We get fragments of text off the web and want to automatically extract structured information from it, which can then be used in other parts of the program.

And in this context I have come across constructions in English which, annoyingly, cannot be defined using context-free grammars. Two examples are:

“The square roots of 16, 9 and 4 are 4, 3 and 2, respectively.”

“The wardrobe has dimensions of 230×120x50 cm (HxWxD).”

In both cases, the problem is that there is a linear matching-up which needs to occur: in the first example, 16 is matched with 4, 9 is matched with 3, and 4 is matched with 2. In the second example, 230 is matched with H (height), 120 with W (width) and 50 with D (depth).

I’m still in the process of figuring out how to best analyse this kind of sentence. Because it’s context-sensitive, a lot of the typical parsing libraries won’t do it out of the box. Which adds to the fun :)

Saturday, 2 August 2008

Something about accessibility…

Written by Johannes Hauser on Saturday, 2 August 2008, 21:55 GMT.
Filed under: electronic devices, mobile, mobile web, usability, user experience.

…which I found in the blog of a german journalist living in London, writing about what happens if you have a great zest for life and are disabled at the same time.

I don’t want to describe it there. Just watch it (it’s a short movie), it’s really worth it. And think about that next time you are designing a web site.

Decision making for experts

Written by Johannes Hauser on Saturday, 2 August 2008, 21:30 GMT.
Filed under: Uncategorized.

Did you know that there is such thing as a World Rock Paper Scissors Society which does hold regular tournaments and country championships and even a world championship (where the world champion will be awarded 10,000 $)?

Maybe this way of decision making is strongly underrated.