[Corpora-List] Suggested Track for Studying Computational Linguistics

[Bob Knippen]

Having studied in both Linguistics and Computer Science departments, I 
think both Chrupala and Line are right. It is fairly easy for a good 
Linguist to understand software but maybe not as easy for a good 
programmer to understand Linguistics. At the same time, it is easier for 
a good statistician to understand enough Linguistics to do statistical 
NLP work than it is for a good Linguist to understand enough statistics 
to do statistical NLP work.

I might even add to this that there is more to computation than 
statistics and programming/software. Many of the formal systems that 
Linguists use (formal language theory, logic, functions/lambda calculus, 
graphs/trees) come from computer science and math and you can learn a 
lot more about them in a computer science department than you can in a 
Linguistics department.

So it seems to me there is no easy answer. You have to make a choice: 
Are you more interested in the 'Computational' part or the 'Linguistics' 
part? It's fairly hard to do both well.

Bob


Mark P. Line wrote:
> Grzegorz Chrupa³a wrote:
> 
>>On 26/09/05, Mark P. Line <mark at polymathix.com> wrote:
>>[...]
>>
>>>And since it's a lot easier for a good linguist to understand
>>>software than it is for a good computer scientist to understand human
>>>language, I'd go for a computational linguistics program that is closely
>>>allied with (or even part of) a linguistics program.
>>
>>FWIW, I think exactly the opposite is true: it is easier to understand
>>linguistics if you're a computing scientist than viceversa. Even
>>though human language is more complex than the subject matter of CS,
>>linguistics is still much less technically (i.e. mathematically)
>>challenging than computing. Compare for example the level of
>>sophistication in chapter 2 (Mathematical Foundations) and chapter 3
>>(Linguistic Essentials) in Manning and Schütze's Foundations of
>>Statistical NLP.
> 
> 
> 
> I don't like that example, because I think statistical NLP is not
> linguistics -- it's what computer scientists do instead of linguistics.
> 
> But there's not really any point in debating the matter here. People will
> make their own decisions about what route to take.
> 
> 
> -- Mark
> 
> Mark P. Line
> Polymathix
> San Antonio, TX
> 
>