On 11/9/11 9:23 PM, Boris Galitsky wrote:
Furthermore it would be nice if you can do the change you did for the>pos tagger
also for the chunker, where you extract the pos tags from the Parse
objects instead
of running the POS Tagger. The Parse object also includes the chunk
information,
so there should be no need to run the chunker.
Hi
I am doing further chunks processing which might be useful for other apps, not
just this 'similarity' project
I need to get all phrases grouped by type (noun, verb, adj, pp, ...) from
chunking results, and it is not clear how can I get phrases other than noun
from 'Parse' object.
Once I get all phrases, I do matching inside my component for each phrase type
separately.
So far I have to process chunking results [1..3 4..5 6...8 6..10] + POS + lemmas
-> lists of phrases for each group.
I suspect there's a better way!
RegardsBoris
Sorry for the late reply. As far as I know it should be possible.
We have code in opennlp.tools.parser.ChunkSampleStream which does it to
train
a chunker based on Parse trees.
Can you try this out and see if it work for you? I guess more people
will need this
anyway, maybe we should create a method somewhere to do this.
Jörn
