[Corpora-List] Corpora for EAP: Architecture...?

Marco Baroni baroni at sslmit.unibo.it
Mon Jan 16 13:18:40 UTC 2006


Hi Eric.

For smallish specialized corpora, I suppose the following Python-based 
solution would work, and it probably would not take more than one day to 
implement...

- Write a script to do random combinations of potentially relevant terms 
from a list

- Use a python module to retrieve web pages from google via the API, e.g.: 
http://pygoogle.sourceforge.net/, using each of the random combinations as 
a query string

- Use the python BTE module (http://www.smi.ucd.ie/hyppia/) to clean the 
pages you retrieve (it's slower than our perl implementation, but for small 
corpora that should not be a problem).

- Use the NLTK or other python/java tools to process the corpus constructed 
in this way

Regards,

Marco



More information about the Corpora mailing list