hooray for Haskell NLP activity!

[Harm Brouwer]

Hey!

On Nov 26, 2010, at 10:04 AM, Eric Kow wrote:

> Woo! It's nice to see this list picking up.
> 

Indeed!

> 1. It's great to see all the tools that people are releasing on hackage.
>   We now have 23 packages listed on
>    <http://hackage.haskell.org/packages/archive/pkg-list.html#cat:natural language processing>
> 
>   What I would love to see now is somebody saying something along the
>   lines of "I built Cool Tool X by combining Y package with Z package"
>   [where Y and Z are built by different teams with different agendas]
> 

That would indeed be cool! We (Daniel de Kok and I) will therefore definitely 
try to release the NLP modules written for our book as general-usuably NLP
libraries.

We might at one point also consider thinking about listing the most valuable
Haskell NLP tools/libs somewhere. Maybe as an appendix.

> 2. The Haskell NLP project could use a maintainer. Anybody interested?
>   Your job is to figure out the Haskell NLP mission is and how to go
>   about fulfilling it.
> 
>   Right now it has a basic mission of building a Haskell NLP community
>   (ie. everybody doing NLP in Haskell should know [a] that other people
>   are working on NLP in Haskell and [b] who they are and what they're
>   doing or [b2] where to find them).
> 
>   Surely you can come up with something closer to world domination?
> 
> 
> 3. Some more Haskell NLP tools: 
> 
> * chartparser (still in progress): general purpose library for building
>  chart parsers.  Comes with interactive debugger so you can poke at the
>  chart.
> 
>     darcs get http://code.haskell.org/nlp/chartparser/
> 
>  I never got a chance to work on this anymore, but maybe in the future.
> 

Cool! A chart parser is definitely something we will also include in the NLPWP 
book. Our current idea is to write an Earley parser. Once we got that nailed down
(time is our worst enemy), I will probably write a Earley-Stolcke modification to it
that allows you to compute prefix probabilities:

http://portal.acm.org/citation.cfm?id=211190.211197

For use in NLP-driven psycholinguistic modeling, like in:

http://acl.eldoc.ub.rug.nl/mirror/W/W10/W10-20.pdf#page=82

** my current implementation is written in Prolog ;-) **

>  I thought it could be handy for building things like teaching libraries
>  ("this is how a chart parser works") or little toy parser prototypes.
> 

That is a nice idea!

> * GenI (hackage, there's a user list you can subscribe and send patches
>  to.  I may have a bit more availability to work on this lately)
> 

>  This is a piece of a natural language generator called a surface
>  realiser. It takes as input a graph of words and a grammar and spits
>  out a list of sentences using the right syntax.
> 
>  Example input:
>    (A) some toy English grammar
>    (B) this graph
> 
>             somewhat
>              |
>              V
>             intense
>              |
>              V
>     john --> love --> mary
> 
>  Example output:
>    John loves Mary somewhat intensely
>    Mary is loved somewhat intensely by John
>    ...
> 
> * fullstop: a laughably stupid sentence segmenter (on hackage)
>  It has some test cases and works if you just want something
>  cheap and cheerful... but really I'm hoping somebody would
>  take over and turn it into something more clever.
> 
>    NLP.FullStop.segment :: String -> [String]
> 
> -- 
> Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
> For a faster response, try +44 (0)1273 64 2905 or
> xmpp:kowey at jabber.fr (Jabber or Google Talk only)
> _______________________________________________
> NLP mailing list
> NLP at projects.haskell.org
> http://projects.haskell.org/cgi-bin/mailman/listinfo/nlp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.haskell.org/pipermail/nlp/attachments/20101126/a338f169/attachment.htm