ctakes-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chen, Pei" <Pei.C...@childrens.harvard.edu>
Subject RE: Using Ctakes takes a long time to process text
Date Mon, 17 Nov 2014 17:25:29 GMT
Budi,
You can also try out ctakes-clinical-pipeline/desc/analysis_engine/AggregatePlaintextFastUMLSProcessor.xml
available in the current 3.2.1-rc.
It contains a new dictionary lookup algorithm from Sean that is roughly 1000% faster for each
pipeline.

--Pei

From: Kim Ebert [mailto:kim.ebert@imatsolutions.com]
Sent: Monday, November 17, 2014 11:55 AM
To: dev@ctakes.apache.org
Subject: Re: Using Ctakes takes a long time to process text

Hi,

cTakes currently is single threaded. To increase throughput, we use a patch or two to run
several pipelines at once. To Increase your performance, I would recommend splitting up the
work into multiple batches.

To get cTakes to run multiple threads, you have to patch the LVG. I believe the patch is available
in the bug tracker.

You also would need to run them as separate pipelines inside of UIMA. You can't just say increase
the number of threads for this operation. While we aren't using static variables, state is
maintained inside of the object. http://uima.apache.org/downloads/releaseDocs/2.1.0-incubating/docs/html/tutorials_and_users_guides/tutorials_and_users_guides.html#ugr.tug.applications.multi_threaded

https://issues.apache.org/jira/browse/CTAKES-151

[IMAT Solutions]<http://imatsolutions.com>
Kim Ebert
Software Engineer
[Office:]801.669.7342
kim.ebert@imatsolutions.com<mailto:greg.hubert@imatsolutions.com>

On 11/16/2014 12:06 AM, Budi Wibowo wrote:

Hello,

i'm using CTAKES for class project.

I'm using CPE to process clinical text notes.

I ran the software with -Xmx10g command.

I have 16g in my machine.



my problem is:

running the CPE takes a long long time.

I'm processing 175 clinical notes with a total of 40MB for all

the notes.

CPE has been running close to 4 hours now,

and it's only been able to process 104 out of

175.



I'm using the "test1.xml" CPE descriptor and

"AggregatePlaintextUMLSProcessor.xml" analysis

engine(clinical-text-pipeline).



 it seems like java is only using 2 core out of the 16 I have. RAM usage hover between 5-8gb.



is there anyway i can make the software run a bit faster?










Mime
  • Unnamed multipart/related (inline, None, 0 bytes)
View raw message