Rasa with spaCy

Rasa Open Source
1

Hello! I have a problem and need help. Rasa 2.8 with spaCy 3.2.0 give a low confidence of intents or mistake in definition of intents For example:

  • intent greed: confidence 0.24
  • intent stop: confidence 0.14 Othes are mistake.

My config.yml, NLU section:

pipeline:
  - name: SpacyNLP
    model: "ru_core_news_lg"
  - name: SpacyTokenizer
  - name: SpacyEntityExtractor
  - name: SpacyFeaturizer
    pooling: mean
  - name: EntitySynonymMapper
  - name: SklearnIntentClassifier
    C: [1, 2, 5, 10, 20, 100]
    kernels: ["linear"]
    "gamma": [0.1]
    "max_cross_validation_folds": 5
    "scoring_function": "f1_weighted"

What I can chenge for high confidence in definition of intents? Please, help

2

Do you mind trying the default pipeline (just for now)?

3

@ChrisRahme, I tryed Spacy with DIET Classifier (without SKlearnIntentClassifier) and it have worked Thank you

4

Glad to know :slight_smile: