About the Statistical models used in Tokenizer / Tagger / Chunker

["ttrung@nlke-group.net" <tt...@nlke-group.net>]

Dear Apache OpenNLP Project Team,

Thank you so much for your amazing tools in NLP field.

I have a question about the statistical models used in these tools:

     - which one (HMM or CRF) did you use to developer the train APIs ?

When I use the Java code train APIs (at 
https://opennlp.apache.org/documentation/1.6.0/manual/opennlp.html) I do 
not see the information about the statistical model.

Thank you so much for your help.

Best regards,

Trung Tran.

Re: About the Statistical models used in Tokenizer / Tagger / Chunker

["ttrung@nlke-group.net" <tt...@nlke-group.net>]

Dear Apache OpenNLP Project Team,

About the previous question, I've checked the models at sourceforge site 
(http://opennlp.sourceforge.net/models-1.5/) and seen that there are two 
models for POS tagging: en-pos-maxent.bin 
<http://opennlp.sourceforge.net/models-1.5/en-pos-maxent.bin> and 
en-pos-perceptron.bin 
<http://opennlp.sourceforge.net/models-1.5/en-pos-perceptron.bin>

Therefore I guess that in the source code for POS Tagger, two models 
were applied: Maximum Entropy Models, perceptron model.

Would you please give me some information about that?

Thank you so much for your help,

Best Regards,

Trung Tran.


On 10/27/2016 06:47 AM, ttrung@nlke-group.net wrote:
> Dear Apache OpenNLP Project Team,
>
> Thank you so much for your amazing tools in NLP field.
>
> I have a question about the statistical models used in these tools:
>
>     - which one (HMM or CRF) did you use to developer the train APIs ?
>
> When I use the Java code train APIs (at 
> https://opennlp.apache.org/documentation/1.6.0/manual/opennlp.html) I 
> do not see the information about the statistical model.
>
> Thank you so much for your help.
>
> Best regards,
>
> Trung Tran.