Understanding Rasa NLU

nahidalam · December 2, 2018, 4:01pm

I am trying to understand how Rasa NLU actually does intent recognition. In multiple previous conversation with @Juste and @alexweidauer , they mentioned that they use BoW approach for intent recognition. I looked into the codebase but looks like they do classification using SVM model and don’t see anything BoW related.

Am I missing something?

naoko · December 2, 2018, 4:58pm

SVM is learning methods and BoW is featurizer. The code to build this BoW featurizer is located here:

github.com

RasaHQ/rasa_nlu/blob/e0f7ec76199968f4856409008aa07064c3f1097f/rasa_nlu/featurizers/count_vectors_featurizer.py#L22


from rasa_nlu.featurizers import Featurizer
from rasa_nlu.training_data import Message
from rasa_nlu.training_data import TrainingData
from rasa_nlu.components import Component
from rasa_nlu.config import RasaNLUModelConfig
from rasa_nlu.model import Metadata


logger = logging.getLogger(__name__)




class CountVectorsFeaturizer(Featurizer):
"""Bag of words featurizer


Creates bag-of-words representation of intent features
using sklearn's `CountVectorizer`.
All tokens which consist only of digits (e.g. 123 and 99
but not ab12d) will be represented by a single feature."""


name = "intent_featurizer_count_vectors"


provides = ["text_features"]

Juste · December 3, 2018, 4:01pm

Hey @nahidalam. Yes, we are using the SVM as a classifier, but the features which are fed into the classifier are created using BoW. Exactly like Naoko suggested already

The simplified process looks like the following:

User input
Bag of words (each token in a sentence gets a word vector assigned)
Sentence representation gets created (by averaging vectors)
It is fed to SVM
The prediction gets produced

faiza_conte · December 19, 2020, 2:07pm

@Juste…i also wanted to know about this things more specificy,can you please explain more of the steps you said here how the user input gets in and processed and prediction get produced…after the bag of words it is confusing me hoe sentence representation gets created , fed to svm and prediction produced…what is SVM tho?

Topic		Replies	Views
Featurization NLU Rasa Open Source	0	466	August 28, 2018
Improve Rasa NLU model Rasa Open Source	5	2016	October 15, 2019
Feedback on ConveRT Model + Rasa NLU Rasa Open Source	59	6359	May 20, 2020
Word Embedding in RASA NLU Rasa Open Source	4	1498	January 14, 2021
Why Rasa NLU uses SVM rather than LSTM? Rasa Open Source	2	3732	April 24, 2019

Understanding Rasa NLU

Related Topics