how should we use the Haskell NLP community?

[David MacIver]

2009/4/28 Eric Y. Kow <eric.kow at gmail.com>:
> Dear Haskell NLPers,
>
> First, welcome to the new subscribers!  Now that we have a bit of mass
> in the list, maybe we should get a discussion started, first on what
> we're trying to accomplish and second, on the kinds of things we can be
> doing as community.
>
> One thing I've noticed is that there actually is a fair bit of work
> going on in the NLP world that's using Haskell, but it all feels a
> little disconnected, as if each of us were the only ones doing this.
> One of the first objectives I had was to build awareness among
> Haskell-NLPers of the other Haskell NLPers.
>
> So there's 15 of us; now what?  What are our goals exactly?
>
> Personally, I happen to like Haskell as a programming language and want
> to be able to use more of it in my NLP work.  My hope is that we can
> make this more likely by building a thriving community: building
> libraries and bindings, supporting each other and NLP folks who are new
> to Haskell, just making Haskell more useful for us in general.  At some
> point, we could even think about how we relate to the wider NLP
> community.  It would be great if we had a clear picture of when exactly
> we can start enthusiastically recommending Haskell as an NLP language
> :-)
>
> What would you like to get out of this community?
> And how do you think we can go about it?
>
> Or maybe another way of looking at the question: what's the one thing
> you could change that would let you use more use more Haskell for NLP?

In the interests of getting some conversation started, I'm going to
use this as an opportunity to soap box slightly. :-)

One thing I feel fairly strongly about is tools taking precedence over
libraries. Libraries are of course important, and many good tools
should be a front end onto a library, but you can write the best NLP
library in the world and it will probably only be used by other people
who care about your language. It might induce some new people to care
about it, but not usually that many. But if you write a tool which can
be used by people who can't even read Haskell then the audience of
people who are interested in it is suddenly dramatically larger.

I also find it useful even if you do know the language the tool is
written in. It forces a useful discipline on you in designing things
(the tools may emerge out of a library, but they force you to
structure the library in a way that allows the tools to be written),
and frequently makes it easier to hack around with quick solutions.

NLP is an area which maps very well onto the unix style of working,
and Haskell is a language which seems to map rather well onto writing
unixy style tools. It would be nice to see people taking advantage of
this.

(Of course I have an ulterior motive in suggesting this, in that I
don't actually write much Haskell these days, so this would make me
much more likely to benefit from the efforts of the Haskell NLP
community. But this is how I work in the languages I do use as well.
:-) )