Logistic Regression classifier does not output proper ranking

[twerkmeister]

Rasa Open Source version

3.2.5

Rasa SDK version

No response

Python version

3.8

What operating system are you using?

Linux, OSX

What happened?

The component does not create a proper intent ranking ... see code. It just picks 10 intents out of all intents (in whatever order they originally were) and then orders them according to confidence...

rasa/rasa/nlu/classifiers/logistic_regression_classifier.py

Lines 141 to 160 in 291f6c6

	def process(self, messages: List[Message]) -> List[Message]:
	"""Return the most likely intent and its probability for a message."""
	X = self._create_X(messages)
	pred = self.clf.predict(X)
	probas = self.clf.predict_proba(X)
	for idx, message in enumerate(messages):
	intent = {"name": pred[idx], "confidence": probas[idx].max()}
	intents = self.clf.classes_
	intent_info = {
	k: v
	for i, (k, v) in enumerate(zip(intents, probas[idx]))
	if i < LABEL_RANKING_LENGTH
	}
	intent_ranking = [
	{"name": k, "confidence": v} for k, v in intent_info.items()
	]
	sorted_ranking = sorted(intent_ranking, key=lambda e: -e["confidence"])
	message.set("intent", intent, add_to_output=True)
	message.set("intent_ranking", sorted_ranking, add_to_output=True)
	return messages

Command / Request

No response

Relevant log output

No response