Get/recognize intents from a csv or similar

199 views
Skip to first unread message

ClauSilva

unread,
Oct 9, 2021, 11:42:25 AM10/9/21
to Dialogflow Essentials Edition users
Hi!  I was wondering if there is a way to use Dialogflow to find which intents exist in a big database.

I mean, I do not want to match with intents I already know, but to find from scratch any intent Dialogflow could find by analyzing  the database, so I can create new flows to answer to those new intents found.

In a practical example: I have a database with all the messages that users have sent to my chatbot and I can filter and get only those that didn't match with my already trained intents. But I have too much "defaults" to analyze manually, so, I was wondering if DF (or another tool) could help me analyzing the text and finding groups?

I was thinking to use DF because that way I can already have trained the intents and build dialogues from there


deva...@kommunicate.io

unread,
Oct 10, 2021, 12:44:12 AM10/10/21
to Dialogflow Essentials Edition users
are you looking to auto create intents from database, or you want to club similar sentences which are not answered by the bot?
For example: if Bot is not able to answer "Take me to Main Menu", "Go to Main Menu", etc. 
As I understand you want to see these grouped together and that will make it easy to decided what intents to create?

Sharing a screenshot below from our product https://www.kommunicate.io/ that does this same thing for Dialogflow and other bots.

Screenshot 2021-10-10 at 10.09.50 AM.png

Sway [Cloud Support]

unread,
Oct 13, 2021, 1:39:05 PM10/13/21
to Dialogflow Essentials Edition users
There currently doesn't seem to be a way to infer unmatched intents using a CSV file (or any other type of file). The only thing that can be done via a CSV file within Dialogflow ES is exporting and importing entities [2].

By default, intent matching is automatically done by the Dialogflow API, and is matched using either "rule-based grammar matching" or "ML matching" [1], which you can further prune using knowledge connectors and context and fallback intents. 

You can submit a request for this feature via the Public Issue Tracker [3], add your email to be notified of all communication which is directly seen by the product development team. Note, that the respective product development team will review the request, which can potentially be implemented at their discretion. Also note that creating a report does not guarantee an eventual implementation.

Reply all
Reply to author
Forward
0 new messages