Deepgram, a company specializing in automatic voice recognition for voice-enabled applications, announced the availability of three new product features: Conversational AI, Sales and Support Enablement and Real-Time Streaming.
The features (real-time streaming, trained and tailored models, multilingual support, on-premise implementation or cloud) have been built, underlined Deepgram, to help enable the next generation of human-like virtual agents and can
In addition, real-time streaming allows software platforms to provide instant sales advice and knowledge-based solutions to sales and support operators.
Most of the existing artificial conversational intelligence solutions, highlights Deepgram, has a lag time of more than two seconds, while the new feature of Deepgram reduces the response delay to <300 milliseconds.
Customers now have a basic model set up on the typical cases of Conversational AI that can be further adapted to specific sectors. These functions allow organizations to improve the error rate of words by up to 50% and reduce noise and filter the crosstalk to help focus on key words and phrases for intent and understanding.
As for Sales and Support Enablement functions, customer-contact teams in each sector can increase sales with an accurate and fast ASR optimized for real-time use.
The characteristics of the new Deepgram product, said the company, help reduce the customer churn with a faster ASR for offers and alerts in real time, and improve the success of employees with a careful analysis of calls and transcriptions for
As for the Real-Time Streaming, Deepgram does not believe that we need to compromise speed to accuracy. With this new processing function, the real-time ASR streaming of Deepgram becomes even faster, with less than 300 milliseconds delay and a transcription accuracy of more than 90%.
In this way, the solutions of artificial intelligence conversational and enabling to sales and support Deepgram, says the company, can now process audio data and provide correct answers, insights and results with customized voice models.