You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by Martin Illecker <mi...@apache.org> on 2015/01/30 11:33:32 UTC

POS Tagging on Storm

Hello,

does anyone have experience with POS Tagging on Storm?

I'm using the Stanford MaxentTagger [1] with the GATE fast model [2] within
a bolt but this is far too slow.
The execution latency of this POS Tagger bolt is around 60 ms and this
slows down the whole topology. (See [3])

Any help would be appreciated!

Thanks!

Best regards
Martin

[1]
http://nlp.stanford.edu/nlp/javadoc/javanlp/edu/stanford/nlp/tagger/maxent/MaxentTagger.html
[2] https://gate.ac.uk/wiki/twitter-postagger.html
[3]
http://mail-archives.apache.org/mod_mbox/storm-user/201501.mbox/%3cCAPbY5A61thA2O6bp4ZOtJV=GY2xHMaaRm5XYjH-zWS=xwvmseA@mail.gmail.com%3e

Re: POS Tagging on Storm

Posted by Martin Illecker <mi...@apache.org>.
No, of course I'm not. By the way the initialization takes up to 4 seconds.
Please have a look [1] [2].

[1]
https://github.com/millecker/storm-apps/blob/master/commons/src/at/illecker/storm/commons/bolt/POSTaggerBolt.java#L81
[2]
https://github.com/millecker/storm-apps/blob/master/commons/src/at/illecker/storm/commons/postagger/POSTagger.java#L62

2015-01-30 12:03 GMT+01:00 Enno Shioji <es...@gmail.com>:

> You are not initialising the tagger on every execution, right?
> ᐧ
>
> On Fri, Jan 30, 2015 at 10:33 AM, Martin Illecker <mi...@apache.org>
> wrote:
>
>> Hello,
>>
>> does anyone have experience with POS Tagging on Storm?
>>
>> I'm using the Stanford MaxentTagger [1] with the GATE fast model [2]
>> within a bolt but this is far too slow.
>> The execution latency of this POS Tagger bolt is around 60 ms and this
>> slows down the whole topology. (See [3])
>>
>> Any help would be appreciated!
>>
>> Thanks!
>>
>> Best regards
>> Martin
>>
>> [1]
>> http://nlp.stanford.edu/nlp/javadoc/javanlp/edu/stanford/nlp/tagger/maxent/MaxentTagger.html
>> [2] https://gate.ac.uk/wiki/twitter-postagger.html
>> [3]
>> http://mail-archives.apache.org/mod_mbox/storm-user/201501.mbox/%3cCAPbY5A61thA2O6bp4ZOtJV=GY2xHMaaRm5XYjH-zWS=xwvmseA@mail.gmail.com%3e
>>
>
>

Re: POS Tagging on Storm

Posted by Enno Shioji <es...@gmail.com>.
You are not initialising the tagger on every execution, right?
ᐧ

On Fri, Jan 30, 2015 at 10:33 AM, Martin Illecker <mi...@apache.org>
wrote:

> Hello,
>
> does anyone have experience with POS Tagging on Storm?
>
> I'm using the Stanford MaxentTagger [1] with the GATE fast model [2]
> within a bolt but this is far too slow.
> The execution latency of this POS Tagger bolt is around 60 ms and this
> slows down the whole topology. (See [3])
>
> Any help would be appreciated!
>
> Thanks!
>
> Best regards
> Martin
>
> [1]
> http://nlp.stanford.edu/nlp/javadoc/javanlp/edu/stanford/nlp/tagger/maxent/MaxentTagger.html
> [2] https://gate.ac.uk/wiki/twitter-postagger.html
> [3]
> http://mail-archives.apache.org/mod_mbox/storm-user/201501.mbox/%3cCAPbY5A61thA2O6bp4ZOtJV=GY2xHMaaRm5XYjH-zWS=xwvmseA@mail.gmail.com%3e
>