Hello wit.ai team,
First of all, thank you for maintaining this project. I truly appreciate the work you put into the platform. I would like to report an issue I’m currently experiencing with incorrect intent classification.
Do you want to request a feature, report a bug, or ask a question about wit?
I would like to report a bug.
What is the current behavior?
I have two intents: greetings and closings.
The greetings intent contains the following utterances:
- good morning
- good afternoon
- good night
The closings intent contains the following utterances:
- see you soon
- bye bye
- catch you later
After completing the training, I tested the intents by sending a message to the wit.ai /message endpoint. However, the response consistently returns the wrong intent, often hallucinating the greetings intent for utterances that clearly belong to closings. Instead of returning the correct intent, the intents field always contains an incorrect one.
If the current behavior is a bug, please provide the steps to reproduce and if possible a minimal demo of the problem.
- Create two intents in wit.ai:
greetings and closings.
- Add the following utterances to greetings:
- good morning
- good afternoon
- good night
- Add the following utterances to closings:
- see you soon
- bye bye
- catch you later
- Train the model.
- Send a message to the wit.ai
/message endpoint using one of the closings utterances (e.g., "see you soon"):
curl \
-H "Authorization: Bearer YOUR_ACCESS_TOKEN" \
"https://api.wit.ai/message?v=20251203&q=see%20you%20soon"
- Observe the response — the intents field always returns the wrong intent:
{
"entities": {},
"intents": [{
"confidence": 0.99762148510878,
"id": "intent_id_here",
"name": "greetings"
}],
"text": "see you soon",
"traits": {}
}
What is the expected behavior?
The expected behavior is that the response should return the closings intent when sending an utterance such as "see you soon". For example:
{
"text": "see you soon",
"intents": [
{
"id": "intent_id_here",
"name": "closings",
"confidence": 0.98
}
],
"entities": [],
"traits": []
}
If applicable, what is the App ID where you are experiencing this issue?
App ID: 340540824156391
Thank you in advance for your help.
Hello wit.ai team,
First of all, thank you for maintaining this project. I truly appreciate the work you put into the platform. I would like to report an issue I’m currently experiencing with incorrect intent classification.
Do you want to request a feature, report a bug, or ask a question about wit?
I would like to report a bug.
What is the current behavior?
I have two intents: greetings and closings.
The greetings intent contains the following utterances:
The closings intent contains the following utterances:
After completing the training, I tested the intents by sending a message to the wit.ai
/messageendpoint. However, the response consistently returns the wrong intent, often hallucinating the greetings intent for utterances that clearly belong to closings. Instead of returning the correct intent, theintentsfield always contains an incorrect one.If the current behavior is a bug, please provide the steps to reproduce and if possible a minimal demo of the problem.
greetingsandclosings./messageendpoint using one of the closings utterances (e.g.,"see you soon"):What is the expected behavior?
The expected behavior is that the response should return the closings intent when sending an utterance such as
"see you soon". For example:If applicable, what is the App ID where you are experiencing this issue?
App ID: 340540824156391
Thank you in advance for your help.