I am working in a research project here at IUSS, Pavia, Italy. Here are some pics from this place. Nice, isn’t it? This is certainly something else than Denmark or Finland. If you happen to visit (this is close to Milano), come to say hello and discuss Finnish with me.
The project I am working on has to do with parsing. I’m using Python (which is a programming language) to see if certain cognitive science/linguistic ideas work or doesn’t work. What I think I am doing is imitating certain aspects of what the brain does when we read/listen to language.
Testing a hypothesis empirically is one thing, but implementing it in a Python code is something different. If you know what programming is like, you already know what I mean. The problem is that the computer understands nothing, so everything has to be formalized to the smallest detail. Every logical step has to be there. On the positive side, programming is very relaxing, because you can always see, with complete rigor, if your ideas work. Here’s an example. Few days ago, I asked the parser to parse the following sentence:
“The claim that the elections were rigged was rejected.”
Simple? You already know what this sentence means? There was a claim that was rejected, and the claim was about elections. But my parser gave the following ‘logical’ solution:
(1) [The claim that the elections] were rigged.
The sentence means some ‘claim was rigged’, and it is not clear what the claim is. The claim that the elections?? The correct parse is this:
The claim that [the elections were rigged]
where it is the elections that were rigged, not the claim. My point is that without this program, it would have never crossed my mind to think that my assumptions would entail (1) as a plausible meaning/parsing of this sentence. Yet it did, because this alternative followed logically from my assumptions. Humans do not think logically. But I like this, because when the computer does something like this, it is always it and not me who is correct. So it’s all like mathematics–unambiguous, logical, clear.
It took me one day and several discussions to fix the issue. It is not trivial, because sometimes you have to do as my parser did in (1), as here:
(2) [The claim from the administration] was rejected.
See? Now it is the ‘claim’, and not the ‘administration’, that is rejected. Why is that? I leave this for you as a home work assignment.
At any rate, I will be posting something about parsing here. The project is not just about writing a parsing application, which would be an applied project, but our main goal is to see if the parsing perspective could help us solving certain open issues in the overall linguistic theory. I have assumed for some time now that this must be the case, and will explain my reasons in a later post.