Copy
Voice on Voice #153 
Tweet Tweet
Share Share
Forward Forward

Building for Inclusivity

Image Source

One of the major pain points of voice assistants is their inability to understand us at times. For those with speech impairments, voice technology has been largely inaccessible. The industry is actively exploring ways to make assistants more inclusive. Last year, startup Voiceitt closed $10 million in Series A funding to build software that can understand those with atypical or impaired speech. Amazon partnered with Voiceitt to improve Alexa. Additionally, Google’s Project Euphonia is analyzing speech recordings to train speech recognition models to understand this community, and Siri has collected 28,000 audio clips with stutters to train Siri. While voice technology for those with speech impairments is still limited, major assistants are actively investing in how they can improve users’ experiences.

Voice Recognition for All

Google introduced Relate, a beta app with voice recognition technology that understands people with speech impairments. The company is aiming to use this data to improve Google Assistant’s speech recognition capabilities and be more inclusive of users that have speech impairments. English-speaking volunteers who sign up for the project will record 500 voice samples to train the AI. Once the recordings are submitted, the users can access features including speech transcription, a feature called “Repeat” that repeats the users’ phrases in a clear voice, and Google Assistant capabilities like controlling smart home devices. By making voice AI more inclusive, the technology can help an audience typically underrepresented in tech.
More on The Verge

New Voices

Nvidia debuted Riva Custom Voice, its toolkit for creating custom synthesized voices with 30 minutes of speech recordings. With the platform, customers can create voice assistants, brand voices, or voice applications. Voice AI synthesis technology is rapidly gaining traction as the time and cost savings are revealed — no longer do companies need to solely rely on having voice actors manually record clips. Companies including Veritone, Lovo, and Sonantic have already deployed their own software for clients to build synthesized voices with their recordings. But as the popularity of custom voices climbs, the industry needs to continue to plan for how these clips might be misused.
More on VentureBeat

Data Discovery

IBM updated its NLP capabilities for Watson Discovery, its enterprise AI service that helps insurance, legal, and financial businesses analyze data from documents. With the NLP updates, the businesses can customize the NLP models to understand industry-specific phrases and terms, better helping the AI glean insights. Voice and conversational AI has tremendous potential for enterprises, especially when it comes to data analysis. Additionally, call centers are already deploying the technology to analyze data from phone calls to better enhance service.
More on Voicebot.ai

By the Numbers

22%

of global smart speaker shipments are from Amazon. (data via S&P Global)

21%

of US internet users will use a smart appliance by 2025, an increase of 10.5% in 2021. (data via eMarketer)

Emerging Tech Stories 

  • Venmo Me. Next year, Amazon will accept Venmo as a payment method on the Amazon website and mobile app.

  • Microsoft x Meta. Meta will integrate Microsoft Teams into Workplace, Meta’s collaborative platform for businesses.

  • No More Writer’s Block. Writer, a startup whose AI-powered assistant helps marketing teams write consistent copy, raised $21 million in Series A funding.
Instagram
Twitter
Facebook
LinkedIn
Making technology invisible, one conversation at a time.
 
Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.