Understanding the ConveyorBot results
Understanding the ConveyorBot layout
When ConveyorBot is confident enough in the matches found in your Knowledge Base, it will draft a proposed answer to the question. In the detail view of the question, you will see a layout that shows the "Input" and "Output":
The "Input" represents the top 5 matches from your Knowledge Base to the question being asked. The "Output" is the generative answer that was drafted after passing those top 5 matches into GPT-3.
If the Output is to your liking, you can hit "Approve and next" to move on to the next question. However, if the answer needs editing, you have a few options:
- Start fresh: Clears the output and gives you a blank canvas to start from.
- Restore input: Reverts the output to the content of your top 5 matches from your Knowledge Base. This is a good way to adjust the input if one of your matches appeared to sully the output.
- Use Output: When the output just needs minor tweaking, this option will let you edit the output directly.
Flagged questions
ConveyorBot will flag questions where it is uncertain how to answer. This may happen for a number of reasons:
- Your Knowledge Base content did not seem to cover the topic at hand, or only obliquely;
- The question was worded in such a novel way that the bot was unable to find the relevant Knowledge Base content
Sometimes, the bot will guess based on the information available, but still flag the question for review.
To remove the flag, click "Remove issue flag" or hit cmd+U
on Mac or ctrl+U
on Windows. Alternatively, you can add an answer then hit "Approve and Next", which will automatically remove the issue flag.
Updated 22 days ago