Topic Detection

Updated on 19 Jul 2024
1 Minute to read
Contributors

Article summary

Did you find this summary helpful?

Thank you for your feedback

A node that accumulates Text based events and detects the topic of accumulated text at the end of operation.

Parameters:

name	description	default
IgnoreSslErrors	Ignore any SSL certificate errors for the Http calls	false
Key	Api key for the Topic Detection service	[enter a valid key]

Inputs

Audio:
none

Events:

name	description	known nodes that generate this event
Every Event that has `Data.Text` field.	i.e SR Milestone events. Translation events, Custom events etc.	SR Http, SR WebSocket, Translate

Outputs

Audio:
none

Events:

name	description
Topic Detected	Contains the top three topics of all the SR Milestone events received in the `Data.Topics` field.

Remarks :

Requirements
The node needs an OpenAI Api key to perform. To learn more about getting an API Key and finding it, you can refer to the OpenAI Api page.

After getting your API Key, you can learn about finding and securely using your key following the instructions on this page.

Project Structure
The node expects events from the SR Http or SR WebSocket node.
A minimum viable project that utilizes the Topic Detection node can be built as such:

Since the Topic Detection node only needs events from the SR nodes, it can be placed wherever an SR node is utilized. A more complicated project that utilizes Topic Detection can be built as such:

Detailed Workflow
Topic Detection node accumulates all the SR Milestone events it receives and waits for a Flush call. (eg. If the client is a Call Recorder application, Flush is typically sent at the end of a call.) When Topic Detection node receives a Flush, it detects the topic of text it had received so far with the help of OpenAI and raises a Topic Detected event.

Supported flow types: Stream, Batch

Was this article helpful?

What's Next

Summarization