[Corpora-List] Help Needed

[Nitin Madnani]

Hello Adeel,

Personally, I don't know of any software that will do the actual
lexical replacement but it shouldn't be hard to code something up
using something like NLTK which comes with WordNet (a database of
synonyms and other lexical relations).

As far as sentential paraphrasing goes, I have written a
broad-coverage sentential paraphraser for my dissertation but
unfortunately it is not publicly available at this time.  Other people
in the community have also tackled this problem before me. However,
again, I don't know of any actual sentential paraphrasing tools that
can paraphrase new sentences.

FYI, if you are interested in actually building a simplified version
of the sentential paraphraser that I have built from my dissertation,
it can easily be done using freely available data and tools. You can
find the details in the two papers below:

(a) Paraphrasing with Bilingual Parallel Corpora. Colin Bannard and
Chris Callison-Burch, 2005. In Proceedings of ACL-2005.

(b) Using Paraphrases for Parameter Tuning in Statistical Machine
Translation. 2007. Proceedings of the Second ACL Workshop on
Statistical Machine Translation (WMT-07). Nitin Madnani, Necip Fazil
Ayan, Philip Resnik, Bonnie Dorr.

Once you read the papers and understand how to construct a sentential
paraphraser, you will need the following data and tools to build the
actual paraphraser:

(a) http://www.cs.jhu.edu/~ccb/howto-extract-paraphrases.html

(b) http://www.statmt.org/moses/

Cheers,
Nitin

PS: I wasn't sure if you'd be interested in this but I thought I would
mention it just in case. Quite a bit of work has also been done on
paraphrase identification as well as the related task of recognizing
textual entailment. You can find more details on those areas at:

[1] http://aclweb.org/aclwiki/index.php?title=Paraphrase_Identification_%28State_of_the_art%29

[2] http://aclweb.org/aclwiki/index.php?title=Textual_Entailment_Portal

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