Has anyone managed to train an NLU pipeline that includes Regex/ entity featurizer component along with XML-Roberta?
If, so I’d like to see the resources that cover this. Thanks.
Has anyone managed to train an NLU pipeline that includes Regex/ entity featurizer component along with XML-Roberta?
If, so I’d like to see the resources that cover this. Thanks.
I’ve done that in the past. It’s mainly just adding the featurizer and using DIET though. Are you referring to that or are you referring to something else?
So you are able to extract intents and entities together for XML-Roberta in this pipeline?
Could you share the pipeline structure for this please? Also, which RASA version did you employ?
I’m assuming you’re asking about a pipeline like this;
pipeline:
- name: WhitespaceTokenizer
- name: "RegexFeaturizer"
- name: CountVectorsFeaturizer
- name: CountVectorsFeaturizer
analyzer: char_wb
min_ngram: 1
max_ngram: 4
- name: LanguageModelFeaturizer
model_name: roberta
model_weights: roberta-base
- name: DIETClassifier
epochs: 100
- name: RegexEntityExtractor
case_sensitive: False
use_lookup_tables: True
yes, so did that work for you? Can it catch intents + entities both?