[Corpora-List] announcing pukwac and wackypedia

[Linas Vepstas]

2010/1/4 Marco Baroni <marco.baroni at unitn.it>:
>> What do you see these being used for? What are the useful applications of
>> dependency-parsed treebanks?
>
> I (and I think many others) got interested in dependency parsers because
> they make it easy to extract tuples that make good features for corpus-based
> semantic models.
>
> For a classic illustration of this approach (in English), see:
>
> Automatic Retrieval and Clustering of Similar Words, Dekang Lin, COLING-ACL,
> 1998, pp. 768-774.

Two follow-ups worth mentioning:
-- Dekang Lin, "DIRT: Discovery of Inference Rules from Text"
   KDD 2001
-- Hoifung Poon & Pedro Domingos "Unsupervised Semantic Parsing"
   Proceedings of the 2009 Conference on Empirical Methods
  in Natural Language Processing}

Both of the above use dependency parses as input.
I believe that other work by Poon, Domingos also uses this for
automated entity extraction, and also for coreference resolution.

In short, when looking for semantic content, dependency parses
seem to provide a simpler, more natural representation of the
syntactic content of a sentence -- they're easier to think about,
manipulate, write code for, etc.

--linas

_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora