Support for users of CCG software closed 7-27-20

[Jonathan Kummerfeld <jkk AT cs.berkeley.edu>]

Hi,
 I think I've got this worked out now - there are a set of features in emnlp8.lbj that map on to what is mentioned in the paper, so I've been looking at various splits of those.

 Next question - is there an easy way to get the system to print the feature vector for each example? I want to be able to see the output of the features for concrete examples (I can look at the code myself, but seeing an actual example is a great way to check my understanding).

Thank you!

Jonathan

On 26 September 2011 13:17, Sammons, Mark <mssammon AT illinois.edu> wrote:

Hi, Jonathan.

I don't believe there is an easy way to specify subsets of features -- as you have surmised, your best chance is via modifying the various .lbj files referred to in the README and then recompiling.
(Possibly, for some configurations you could compile with only a subset of the .lbj files, but I don't know how the feature specification is organized with respect to the published results.)

Regards,

Mark

---

From: illinois-ml-nlp-users-bounces AT cs.uiuc.edu [illinois-ml-nlp-users-bounces AT cs.uiuc.edu] on behalf of Jonathan Kummerfeld [jkk AT cs.berkeley.edu]
Sent: Monday, September 26, 2011 3:11 PM
To: illinois-ml-nlp-users AT cs.uiuc.edu
Subject: [Illinois-ml-nlp-users] Question about reproducing the results in table 6 of the EMNLP 2008 paper

Hi,
 I am interested in reproducing the coreference resolution data (numbers and output) that was used for table 6 in Bengston and Roth (2008). I've downloaded the relevant packages and built the coreference system, but I can't see how to run anything other than the full system. Is there a simple way to run with the various subsets of features? Or do I need to adjust the code, removing features by adjusting emnlp8.lbj?

Thank you!

Jonathan Kummerfeld
Graduate Student - Computer Science, UC Berkeley