Training Portal: The Discovered Queue
What does the discovered queue do?
The discovered queue helps give you insight into the questions that your teammates or customers have asked when Talla does not have an answer, but a training request was not created.
The discovered queue is populated from a few different streams:
By default, the discovered queue is populated with queries that meet each of the following criteria:
- a user asks a query that Talla does not have an answer to (i.e. Talla responds with "I didn't find any results...do you want to open a request?"
- the user does not open a training request
- the query must also look like a legitimate question
- for this criteria, we apply natural language processing (NLP) to determine whether or not the user's query looks like a legitimate question. This helps remove clutter that may be caused by queries that contain only one or two words, or that may be irrelevant to your knowledge base.
Queries from Slack
Admins can configure the discovered queue to look at specific Slack channels to highlight the questions your team is asking for you to train Talla to answer. Slack can be configured to extract queries from Slack in the following ways
Historical queries in Slack
Select this option to extract historically asked questions from a specific Slack Channel. Talla can extract questions from the last 1, 3, or 6 months of conversational history
Recurring queries in Slack
Select this option for Talla to continuously listen to a specific Slack channel and surface relevant questions that your team is asking about
Queries from CSV
Select this option to upload a CSV containing commonly asked questions. This gives you the opportunity to test out and train your Talla bot's responses for a specific set of questions.