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

["Sammons, Mark" <mssammon AT illinois.edu>]

Hi, Nathan.

I'll take a look; probably a first step is to add a bunch of logging 
statements to get a better sense of what is failing to happen.  It may take 
us a while to work through this, so I'm going to offer an alternative:
We do have a new Java version of SRL, which depends on our "Curator" 
software.  We're currently working on an improved version of the Curator 
software, and are planning to release it by the end of the summer.  But we 
also have a working version of the Curator + SRL that you could try if you 
are willing; installation is not yet as straightforward as we'd like. 

If you want to try this current version, you can get it here:

http://cogcomp.cs.illinois.edu/trac/wiki/Curator

-- this page also has some documentation for Curator.  The bootstrap.sh 
script that you run as part of the installation process will download SRL and 
other UI NLP tools that may interest you.  You don't have to run them all, 
and Curator allows you to run different NLP tools on different machines as 
some of the tools are quite memory hungry.  But you could just run the SRL 
and its dependencies (which should be clear from the Curator config files)
on a single machine if it has 8G of RAM or more. (This is playing it pretty 
safe).

You may also want to use Edison, a java library that allows more 
straightforward interactions with the Curator server and provides a lot of 
NLP data structures and supporting functionality, which is here:

http://cogcomp.cs.illinois.edu/page/software_view/Edison

-- when you agree to the academic license, it will show a page with some 
introductory material for using Edison.

Please let me know what you decide. 

Regards,

Mark


________________________________________
From: 
neatnate AT gmail.com
 
[neatnate AT gmail.com]
 on behalf of Nathan Schneider 
[nschneid AT cs.cmu.edu]
Sent: Monday, July 23, 2012 4:15 PM
To: Sammons, Mark
Cc: 
illinois-ml-nlp-users AT cs.uiuc.edu
Subject: Re: [Illinois-ml-nlp-users] srl-demo2 hanging

Mark,

With that sentence it's the same except in the logs, fexidentification
has "Processing..." 3 times and fexclassification has "Processing..."
2 times.

Nathan

On Mon, Jul 23, 2012 at 2:08 PM, Sammons, Mark 
<mssammon AT illinois.edu>
 wrote:
> Hi, Nathan.
>
> Do you get the same result with a longer sentence -- say, "John said he 
> read the book and laughed."?
>
> Regards,
>
> Mark
>
> ________________________________________
> From: 
> neatnate AT gmail.com
>  
> [neatnate AT gmail.com]
>  on behalf of Nathan Schneider 
> [nschneid AT cs.cmu.edu]
> Sent: Monday, July 23, 2012 3:53 PM
> To: Sammons, Mark
> Cc: 
> illinois-ml-nlp-users AT cs.uiuc.edu
> Subject: Re: [Illinois-ml-nlp-users] srl-demo2 hanging
>
> Mark,
>
> Thanks for the reply. That does indeed solve the hanging problem, so
> the tool now terminates—but unfortunately it does so without printing
> any result. Here is what I'm trying:
>
> $ echo "Misery loves company ." | ./srl-client-primitive.pl 0 0
>
> The output in the logs is as follows:
>
> ==> charniak-server.log <==
> [Connect from LOCALHOST]
> Parse: <s> Misery loves company . </s>
> Parse ok
> [Connection closed from LOCALHOST]
>
> ==> classification.log <==
> [Connect from LOCALHOST]
>
> ==> fexclassification.log <==
> fex: got connection from 127.0.0.1
> Waiting for clients...
> Processing...
>
> ==> fexidentification.log <==
> fex: got connection from 127.0.0.1
> Waiting for clients...
> Processing...
> Closing client socket...
>
> ==> identification.log <==
> [Connect from LOCALHOST]
> [Connection closed from LOCALHOST]
>
> ==> snowclassification.log <==
> SNoW: got connection from 127.0.0.1
> Waiting for clients...
> Options: -o allactivations
>
> ==> snowidentification.log <==
> SNoW: got connection from 127.0.0.1
> Waiting for clients...
> Options: -o allactivations
> Closing client socket...
>
> ==> srlservercentral.log <==
> [Connect from LOCALHOST]
>
> ==> classification.log <==
> [Connection closed from LOCALHOST]
>
> ==> fexclassification.log <==
> Closing client socket...
>
> ==> snowclassification.log <==
> Closing client socket...
>
> ==> srlservercentral.log <==
> [Connection closed from LOCALHOST]
>
>
> Any suggestions for how I should go about debugging this?
>
> Cheers,
> Nathan
>
> On Mon, Jul 23, 2012 at 11:51 AM, Sammons, Mark 
> <mssammon AT illinois.edu>
>  wrote:
>> Hi, Nathan.
>>
>> Please try the following: in bin/inference.pl on line 107, you should see 
>> the line:
>>
>> my $search_mode = 2;
>>
>> please change it to
>>
>> my $search_mode = 1;
>>
>> and retry.  Please let me know if this fixes the problem.
>>
>> Regards,
>>
>> Mark
>>
>> ________________________________________
>> From: 
>> illinois-ml-nlp-users-bounces AT cs.uiuc.edu
>>  
>> [illinois-ml-nlp-users-bounces AT cs.uiuc.edu]
>>  on behalf of Nathan Schneider 
>> [nschneid AT cs.cmu.edu]
>> Sent: Friday, July 20, 2012 2:47 PM
>> To: 
>> illinois-ml-nlp-users AT cs.uiuc.edu
>> Subject: Re: [Illinois-ml-nlp-users] srl-demo2 hanging
>>
>> Hi,
>>
>> I am trying to use the SRL system and am encountering the same issue
>> (it hangs during beam search). Has a solution been found?
>>
>> Thanks,
>> Nathan
>>
>> http://lists.cs.uiuc.edu/pipermail/illinois-ml-nlp-users/2011-May/000090.html
>>
>> On Wed May 11 19:51:58 CDT 2011, Steven Corman <steve.corman at asu.edu>  
>> wrote:
>>> Greetings.  I am trying to get the semantic role labeler running.  I
>>> think I have got everything compiled and configured properly, but when
>>> I try to process a single, simple tokenized sentence, it hangs at some
>>> point along the way.  It appears from the log files (contents below)
>>> that all of the components with logs are working as they should.  The
>>> problem seems to be in the bin/beamsearch/search executable.
>>>
>>> As you can see from the top iteration captures (below the log
>>> contents, showing only processes using CPU to save space)  at the
>>> second iteration several srl components are running and a zombie
>>> appears.  By the next iteration, search, fex, and snow are running.
>>> By the last it's just search, consuming almost all of one cpu (out of
>>> two).  It sits there like that ad infinitum.  If I kill the process
>>> holding search, inference.pl comes to the top, runs for a while, then
>>> the job ends with no output.  The /tmp/2127.result file mentioned in
>>> the srlservercental log is filled with several thousand lines that
>>> say: double beam!!
>>>
>>> Anyone have a suggestion for what might be wrong, or how to debug?
>>>
>>> Cheers,
>>>
>>> Steve
>> _______________________________________________
>> illinois-ml-nlp-users mailing list
>> illinois-ml-nlp-users AT cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/illinois-ml-nlp-users