- To encourage adoption, the companies have released all the materials used to generate the voice to the Open Source community
- This year, consumers interacted with 4.2 billion digital voice assistants around the world, and that number is expected to double to 8.4 billion devices by 2024, according to a report by Juniper Research
Accenture has collaborated with CereProc, a text-to-speech technology provider, to create Sam, a non-binary voice solution for the fast-growing global digital assistant market. To encourage adoption, the companies have released all the materials used to generate the voice to the Open Source community.
This year, consumers interacted with 4.2 billion digital voice assistants around the world, and that number is expected to double to 8.4 billion devices by 2024, according to a report by Juniper Research. The vast majority of today’s device voices are female, or female by default, and this limited diversity is problematic.
A 2019 UNESCO report found that designing female-only voice assistants reinforces gender bias and encourages negative behavior, both with digital assistants and with real people. Additionally, voices available today are binary and do not reflect the transgender or gender non-conforming community, which represents 12 per cent of millennials.
Text-to-speech model using its artificial intelligence (AI) technology
To help solve these issues, researchers at Accenture Labs worked closely with members of the non-binary community on the development of Sam’s voice. Accenture surveyed non-binary individuals and used their feedback and audio data to influence not only pitch, but speech patterns, intonation, and word choice. CereProc then created the text-to-speech model using its artificial intelligence (AI) technology. The result is a voice that combines aspects of male and female voices to better resonate with the community it was designed to represent.
Marc Carrel-Billiard, senior managing director and Technology Innovation lead at Accenture said, “While gender-neutral voice samples have been released previously, Sam is the first non-binary AI-based digital voice solution that can be embedded into any software solution to speak text in a human-sounding voice. This work is a great example of how technology and human creativity can come together and spark a moment of societal change that can benefit many people. We believe it’s essential that voice assistants more accurately represent the diversity of our global population.”
Open-source engine, along with voice-training data
Accenture has open-sourced all components necessary to develop a non-binary voice assistant. It includes a version of the text-to-speech voice running on an open-source engine, along with voice-training data to encourage broader adoption. Accenture is also working with researchers at Heriot-Watt University who will use the voice as part of a collaborative research effort with several other universities focused on designing conversational assistants to reduce gender bias.
Paul Welham, chief executive officer of CereProc, added, “By creating a non-binary voice with Accenture, CereProc is challenging strongly held mainstream views in IT that a synthetic voice has to be clearly male or female. One of CereProc’s key aims since inception has been to empower application designers to disrupt the status quo in speech; with this non-binary voice we wish to raise awareness of this important issue in the next generation of AI-based digital voice systems that are developed.”