SynCog gives voice agents real-time signals about how people communicate, alongside what they say, helping agents adapt their responses to each person’s spoken communication style and context.
Speech-to-text keeps the words. It loses how people think, connect and express themselves when they speak.
Experts score real speech
Communication specialists rate over twenty speech variables, from pace and prosody to reasoning, word choice, turn-taking, active listening and receptiveness to other views.
Validity and reliability
Through controlled experiments we test inter-rater agreement, test-retest reliability and construct validity to build our gold standard.
Models learn to do the same
Our models learn to score speech the way the specialists do, and are tested on people they have never heard.
Signals in real time
During a live conversation, SynCog passes these signals to the voice agent, so it can adjust its response to what the person says and how they say it.
Every signal SynCog produces maps to one of four strands of spoken communication.
Every signal is anchored to how trained specialists judge spoken communication, and tested before it is used.
