Twitter makes public conversations more accessible with Microsoft AI

Everyone knows Twitter. As of January 2022, it’s where over 229 million daily active users go to see and talk about what’s happening. In 2020, Twitter introduced Spaces, a completely new way to bring conversations on Twitter to life. Spaces expanded the public conversation with live audio, but also presented an opportunity to ensure its experience is accessible to everyone. 

To help, Twitter partnered with Microsoft to build a custom speech-to-text solution using Azure Cognitive Services that adds live automated captions to Spaces. Now, Spaces can reach — and connect — more people than ever before.

A more intimate space

Launched in 2020, Spaces is a public platform that supports ephemeral, live audio conversations on Twitter Led by people’s voices, Spaces aims to help people feel more connected while maintaining the emotion, nuance, and empathy that can be lost in text. By going beyond a character limit, people can have extended and more intimate conversations on Twitter.

When a host creates a Space, anyone can join as a listener. The host can then designate other speakers, and suddenly a conversation is in full bloom.

Connecting a diverse world

“Serving the public conversation means continuing to take the right steps to make Twitter accessible to all,” says Nick Caldwell, General Manager of Core Tech at Twitter. ”We’ve been working in the open to make Spaces more accessible for people with disabilities. Partnering with Microsoft Azure Speech-to-Text allows us to accelerate our progress.” 

When Twitter launched Spaces, the team openly worked to identify and address accessibility gaps in the product, such as improving the accuracy and synchronization of captions with audio and options for increased language support.

Keeping public conversation accessible

From the moment Spaces was first launched to a small feedback group, Twitter openly shared updates and insights based on people’s responses. Captions are important for Spaces, and their teams made it a priority to make them as accurate and versatile as possible. To help, Twitter partnered with Microsoft to develop a solution that could provide high-quality, accurate captions.     

Introducing auto captions

Speech-to-text (also known as speech recognition) from Azure Cognitive Services enables real-time transcription of spoken audio into text. The service supports more than 100 languages and variants, defaulting to a Universal Language Model deployed on Microsoft Azure. The service can even be customized and tailored to different acoustic, language, and pronunciation models to handle ambient noise or industry-specific terminology.

Every Space now automatically comes with live captions, and listeners can choose to see captions just by tapping a button in the Spaces settings menu. 

Making the future more accessible

Twitter sees automated captions as part of a long-term investment in accessibility. In addition to auto-captions, Spaces was released with support for screen readers, and the company has also released a sensory guide that identifies stimuli people may encounter in Spaces, like sound and motion levels.

By partnering with external groups and gathering feedback from users, especially those within the disabled community, Twitter will continue fostering public conversations while striving to make its platform welcoming and accessible for all.

Learn more about Twitter's commitment to accessibility and inclusion by watching a panel on captioning from the 2022 Microsoft Ability Summit.