I actually have a bot that use lookup table. That’s why I need to use the
CRFEntityExtractor to extract my entities.
My intents are quite unbalanced, that’s why I would like to use the
batch_strategy: sequence of
DIETClassifier to help with this.
Here is my Pipeline:
language: "fr" pipeline: - name: WhitespaceTokenizer - name: RegexFeaturizer - name: LexicalSyntacticFeaturizer - name: CountVectorsFeaturizer - name: CountVectorsFeaturizer analyzer: "char_wb" min_ngram: 1 max_ngram: 4 - name: "CRFEntityExtractor" - name: DIETClassifier batch_strategy: sequence epochs: 50 - name: EntitySynonymMapper
The CRFEntityExtractor detect correctly the entities, however the DIETClassifier will not detect the correct intent (because it doesn’t know the entities from the lookup table)
Is there a way to specify which entity extractor we want to use to the DIETClassifier ? If not, is there anyone that have an idea how to handle unbalanced Intents and lookup tables ?