All the latest updates about Answerly & Products. You can now enable or disable several rules for answer quality control!
We support the following Quality Control rules:
- Prevent Hallucinations
- Prevent Unrelated Conversations
- Prevent Offensive Messages
You can access this option in the new Quality Control settings page.
Prevent Hallucinations
This system checks each query against your Knowledge Hub to determine a confidence score.
For queries scoring low on confidence, the chatbot suggests it’s unsure to answer, and usually directs the user to the contact form.
Prevent Unrelated Conversations
This system checks queries to see if they are unrelated to your business and the Knowledge Hub.
If the query is unrelated, the chatbot responds that it can’t help with that query.
Prevent Offensive Messages
This system checks queries using OpenAI’s Content Moderation API.
If the query contains vulgar content or violates OpenAI’s ToS, the chatbot will respond with a default message: “I can’t help with that.”
Good to Know
Hallucinations and offensive messages are enabled by default. Unrelated conversations are disabled.
All these systems use additional tokens, and may slow down the bot responses by 1 second per activated system.
If our quality control systems kick in, your custom prompts will be overriden!