Encountered an exception while running action ‘train_form’. Bot will continue, but the actions events are lost. Please check the logs of your action server for more information.
Above is the error I got in the middle of the conversation. Actions file is fine and under ‘rasa run actions’ command I got errors stating "Failed to extract slot “some_x” with action “some_y”.
where did it went wrong.
Hi @ManiNuthi, that means that it wasn’t able to extract the slot from the input you gave. Most likely it is because the intent was classified as something you weren’t expecting, an entity wasn’t picked up, it failed validation, etc. I’d urge you to run both servers on --debug mode to get more information.
This model works fine for english, but fails to classify intent properly for other languages( in my case telugu). Is it because of rasa’s latest version or SKlearn Intent classifier ?
It could be for a number of reasons, but yes that sounds like an NLU problem then. I wouldn’t recommend the Sklearn intent classifier, but rather this pipeline from here
If there are no word embeddings for your language or you have very domain specific terminology, we recommend using the following pipeline:
language: "fr" # replace with your two-letter language code
pipeline:
- name: WhitespaceTokenizer
- name: RegexFeaturizer
- name: LexicalSyntacticFeaturizer
- name: CountVectorsFeaturizer
- name: CountVectorsFeaturizer
analyzer: "char_wb"
min_ngram: 1
max_ngram: 4
- name: DIETClassifier
epochs: 100
- name: EntitySynonymMapper
- name: ResponseSelector
epochs: 100
No this language do have standard embeddings. May be I have to modify intents and try again.
The pipeline above does not use any embeddings, it learns on your data directly.
does data means intents here. Then I need to have lots of samples in each intent.
Yes, and you need to make sure that the intents are also structured well, i.e. with not much crossover between intents, etc