Now into our third and final year of the Nuvoic Project, we’re seeing exciting progress in Voiceitt’s work on automatic speech recognition for atypical speech, and significant developments on their new dictation and captioning tool. Results are at an early stage, but they look promising!
Continuous recognition of atypical speech
Voiceitt’s app already supports people with atypical speech to communicate and control Smart Home technologies more easily, but users currently have to train each phrase in advance. In the next release, Voiceitt aims to offer real-time recognition, allowing people with non-standard speech to have their words transcribed or interpreted, without the need to pre-train every phrase.
To achieve this Voiceitt have launched Project Ensemble, a website where volunteers can donate recordings of atypical speech patterns as training data for Voiceitt’s recognition models. Through Nuvoic, we’ve been recruiting participants, offering support to take part and gift cards as a thank you for all contributions. So far we’ve supported 28 people to join Ensemble, who have uploaded an amazing 15,000 phrases in total – congratulations and thank you to everyone who’s contributed so far! 🤩🎉
Voiceitt are already using these and recordings from other participants around the world to develop a new dictation and captioning tool. In the video below, Michael demonstrates his personal speech model providing live captions through Zoom (Zoom’s built-in captioning doesn’t work for Michael’s speech). Users can also dictate text for use in messages or documents, or to share onscreen with a conversation partner. While this technology is still at an early stage it’s a significant improvement on what’s currently available, and with more recordings will become more accurate over time.
However, we still need more participants to donate recordings to help develop this technology more quickly and test the results. We want to include as many non-standard voices as possible so please get in touch if you or people you support would like to get involved. We can offer equipment on loan, technical support, gift cards of up to £100 for participants and funding for organisations and individuals supporting someone to take part. If you or your organisation would like to be involved in this exciting development work, please see our web pages or contact our project co-ordinator: liz@karten-network.org.uk for more information.
Feedback from users
As well as their work on continuous recognition, Voiceitt have continued to use feedback from our participants to improve their app. Recent additions include: a button to flag misrecognitions, helping to improve accuracy; the option to repeat spoken output; and more options to use Bluetooth accessories in combination with the internal mic and speaker, to support the use of Voiceitt ‘out and about’ or in noisy environments. All of these features have been requested by users and it’s great to see them now available!
Other Updates
In September we took Ensemble on the road, exhibiting at the Communication Matters conference in Leeds. We had a lot of interest in the project, which we hope will generate some new recruits!
Our Project Technologist, Geena, recently contributed an article to the Jisc Accessibility and Assistive Technology blog – thank you, Geena! Read Geena’s perspective on How can we make speech recognition technology work for everyone, including those with the most significant speech impairments?
Get in touch!
We would love to hear from you if you, or someone you know may be interested in taking part in donating speech samples or testing Voiceitt’s technology. Please email our project co-ordinator: liz@karten-network.org.uk, or you can find more information and get in touch via our project web pages.
Article meta data
Clicking on any of the links in this section will take you to other articles that have been tagged in the same category.
- Featured in the Karten Autumn 2022 Newsletter
- This article is listed in the following subject areas: Nuvoic Project