Hi,

I actually would not solve this task with Ruta, but would rather acquire
more training data and apply single-label multi-class classification
using a linear svm with some iterations for the feature
extraction/weighting/normalization... as a start...

If you need to approach this with Ruta, I would do some simple
dictionary lookup for the keywords and apply some postprocessing like:

DECLARE Intent (String value);
WORDTABLE intentTable = 'intent_table.csv';
MARKTABLE(Intent, 1, intentTable, true, 4, ".,-", 2, "value" = 2);
Intent->{ANY i:@Intent{-> UNMARK(i)};};

with the table looking like:

new credit card application;Apply_for_Card
I want to apply for a new card;Apply_for_Card
papers needed to apply for a card;Apply_for_Card
open a MasterCard;Apply_for_Card
open a new credit card;Apply_for_Card
card application;Apply_for_Card
application for a new personal credit card;Apply_for_Card-Personal
I want to apply for a new personal card;Apply_for_Card-Personal
Open a business card;Apply_for_Card-Business
balance;Balance-Inquiry
...


(Param dictRemove WS = true)



Best,

Peter



Am 11.04.2018 um 19:57 schrieb Igor Mayer:
> Hello, Peter!
>
> Hope you are doing well. 
> I am a new user of UIMA RUTA, and sorry, that I dare to ask you
> questions directly, but I have seen this email address at
> StackOverflow and you said there, that it takes less time normally to
> receive an answer. I have to solve the exercise attached to this
> letter. I fully read the UIMA RUTA Guide & References posted on the
> Apache.org. However, I didn't found the good approach there.  
> I have a task, the script has to 'understand' each sentence\utterance
> (one by one) and link each to one of the intents. I tried to use
> Regular Expressions as keywords, and sort sentences with some keywords
> with Contains statements. However, this approach looks really
> duplicative. So, would you be kind to help me, may I ask you to tell
> the best approach, to solve this task and an example of how it should
> look like?
>
> Thank you a lot in advance! 

-- 
Peter Klügl
R&D Text Mining/Machine Learning

Averbis GmbH
Tennenbacher Str. 11
79106 Freiburg
Germany

Fon: +49 761 708 394 0
Fax: +49 761 708 394 10
Email: peter.klu...@averbis.com
Web: https://averbis.com

Headquarters: Freiburg im Breisgau
Register Court: Amtsgericht Freiburg im Breisgau, HRB 701080
Managing Directors: Dr. med. Philipp Daumke, Dr. Kornél Markó

Reply via email to