Hi everyone, I am building a Rasa Bot the first part of the conversation relies on Buttons only. I fill my slots according to the button clicked, which works fine. But the conversation often gives the wrong answer or jumps in a wrong story. I suspect this has something to do with the training? Afterall since I’am only using buttons I do not have a lot of trainings data for my intents. Is that possible? Or do I need to adapt my pipeline in order to achieve the correct outcome. I would appreciate your input.
Hi. Maybe you can show your stories and domain file.
Your config.yml file would help too.
Unfortunatly I can not share because it is a client project. But I got it working now…setting the max_history to 0. I will share my pipeline, nothing proprietary about that language: de
pipeline:
- name: "SpacyNLP"
model: "de_core_news_sm"
- name: "SpacyTokenizer"
- name: "SpacyFeaturizer"
- name: "RegexFeaturizer"
- name: "CRFEntityExtractor"
- name: "DIETClassifier"
epochs: 200
policies:
- name: AugmentedMemoizationPolicy
max_history: 0language: de
pipeline: - name: “SpacyNLP” model: “de_core_news_sm” - name: “SpacyTokenizer” - name: “SpacyFeaturizer” - name: “RegexFeaturizer” - name: “CRFEntityExtractor” - name: “DIETClassifier” epochs: 200
policies: - name: AugmentedMemoizationPolicy max_history: 0