Is it possible to check all the classified intents and their respective entities? Currently, the model outputs the result with the highest confidence. How can I query all the other intents one by one regardless of confidence and check for predictions?
Hi @sumitsaha, you can use the HTTP API to parse a message with your Rasa model and get the full intent breakdown.
The response object contains the element intent_ranking
that stores all the confidence information:
{
"text": "Hi",
"intent": {
"id": 3351123226924298327,
"name": "greet",
"confidence": 0.9998955131
},
"entities": [],
"intent_ranking": [
{
"id": 3351123226924298327,
"name": "greet",
"confidence": 0.9998955131
},
{
"id": -7503364390322699277,
"name": "chitchat",
"confidence": 0.0000701769
},
{
"id": -7477535035922634779,
"name": "thankyou",
"confidence": 0.0000227083
},
{
"id": -1399674655192123369,
"name": "bot_challenge",
"confidence": 0.000004893
},
{
"id": -1208978872258037789,
"name": "stop",
"confidence": 0.0000036492
},
{
"id": 8816719142779407201,
"name": "deny",
"confidence": 0.0000020244
},
{
"id": -1750768160529046574,
"name": "affirm",
"confidence": 0.0000009217
},
{
"id": 4137433111448671806,
"name": "inform",
"confidence": 0.0000000938
},
{
"id": -4386500614336766294,
"name": "request_restaurant",
"confidence": 0.0000000004
}
]
}
1 Like