US 12,314,663 B1
Endpoint detection
Adam Earle, Palo Alto, CA (US); Ali Ziaei, Palo Alto, CA (US); Jack Weissenberger, Palo Alto, CA (US); and Romain Cosentino, Palo Alto, CA (US)
Assigned to Salesforce, Inc., San Francisco, CA (US)
Filed by Salesforce, Inc., San Francisco, CA (US)
Filed on Apr. 16, 2024, as Appl. No. 18/636,671.
Int. Cl. G06F 40/284 (2020.01)
CPC G06F 40/284 (2020.01) 20 Claims
OG exemplary drawing
 
1. A method comprising:
generating a log of an ongoing audio conversation in real time that involves at least a first user, wherein the log comprises a plurality of tokens that represent text data generated in real time via transcription of the ongoing audio conversation, wherein participants in the ongoing audio conversation have taken turns communicating during the ongoing audio conversation;
receiving, using a large language model (LLM), a context of the ongoing audio conversation comprising at least one token from the plurality of tokens;
predicting, using the LLM and the context, a next token;
predicting, using a classifier and the next token, a completion of a current turn of the first user speaking; and
triggering a next turn of the ongoing audio conversation in real time using the completion of the current turn.