Intent coming based on training example

i intent to pass user message through rasa to get message intent and entity but intent coming based on training examples i dont want that below is my configuration for config.yml file pipeline: - name: “SpacyNLP” model: “en_core_web_md” case_sensitive: false - name: “SpacyTokenizer” - name: “SpacyFeaturizer” - name: “SpacyEntityExtractor” - name: “SklearnIntentClassifier” C: [1, 2, 5, 10, 20, 100] kernels: [“linear”] output i am getting after training messages is { “intent”: { “name”: “affirm”, “confidence”: 0.29226867387012845 }, “entities”: [ { “entity”: “GPE”, “value”: “India”, “start”: 9, “confidence”: null, “end”: 14, “extractor”: “SpacyEntityExtractor” } ], “intent_ranking”: [ { “name”: “affirm”, “confidence”: 0.29226867387012845 }, { “name”: “bot_challenge”, “confidence”: 0.2919755203435614 }, { “name”: “greet”, “confidence”: 0.13382467018268535 }, { “name”: “deny”, “confidence”: 0.07883030355784834 }, { “name”: “mood_unhappy”, “confidence”: 0.07520340386823511 }, { “name”: “goodbye”, “confidence”: 0.07368384154623404 }, { “name”: “mood_great”, “confidence”: 0.054213586631306904 } ], “text”: “Where is India” } i dont want intent based on training messages is it possible i am new to RASA any help would be appreciated :innocent:

Maybe I’m misunderstanding what you mean, but the whole idea with an NLU model is to classify input based on training data. Could you post an example of your NLU data, and your input and expected output?

Fyi: You can use markdown formatting in your replies to maintain your code formatting. Either use backticks e.g. `code` or indent a paragraph by 4 spaces e.g.

i am
a multiline
code block

config.yml (773 Bytes)

Above i have attached file with config.yml config .is it possible to get intent of user message without providing predefined intents like if user write Where India is located intent should be something like “find_india” .How can i achieve that or is it compulsary to provide predefined intents

Ok, I understand what you’re asking now. In short, no, that’s not possible in Rasa at this point. Rasa works using predefined intents. Even if you built a custom component that could predict intents without having seen them, the stories would still need to be written using predefined intents to know what action to take.

You might find this blog post interesting: It’s about time we get rid of intents