What does a publisher sound like? Meet Amy, the FT voice
‘Clear, steady and calm’: Meet Amy, the voice behind the FT
By Ella Wilks-Harper | May 18, 2018
More publishers are tuning in to the exciting developments happening in voice technology. Reuters Institute research found that 58% of publishers are focusing on podcasts and content for voice activated-speakers.
With the likes of Amazon Echo and Google Home taking off rapidly, publishers are working on what they will sound like in future.
For the last two years, FT Labs, a small technology team at the British newspaper, have been working to answer this question.
“The actual quality of voice has increased so much that we can now contemplate putting it in front of our subscribers” Chris Gathercole, head of FT Labs, told Reuters. “Up until we started experimenting with voice, there was no such thing as the FT voice. We don’t have a brand team that says the FT is baritone from northeast of London” he added, half-laughing.
Up until we started experimenting with voice there was no such thing as the FT voice. We don’t have a brand team that says the FT is baritone from northeast of London”
Lessons learnt so far
Despite only being in its early stages of implementation, there is promise that the option to listen to articles is something some subscribers want.
Gathercole noted that within the 10% of subscribers who have the option, there is a small percentage who actively use the audio option multiple times each month.
Currently, Amy is pumped through a text-to-speech translation engine and speaks out every literal word of 400 to 500 articles each day.
I would imagine (within five years time) it would be possible to switch between audio and text and video across multiple devices. So whichever one you happen to be near at the time will carry on the story”
Amy’s obstacles
However, the team soon discovered that each article needs to be stripped of metadata and arbitrary text structures in order for the technology to work. This poses a potentially looming task if the team later decide to incorporate audio for all previously written articles.
Another conundrum facing the team is Amy’s compatibility with the FT’s style guide.
Particular phrases and countries have been amended in order for Amy to properly translate them, such as North Korea, that is written in the FT as ‘N.Korea’.
Even saying the ‘FT’ initially proved a challenge that the team managed to amend by programming.
Despite these initial teething problems, Amy is creating a buzz in the newsroom. Last year, FT journalist Lucy Kellaway wrote about Amy noting that though Amy might not have the same human intuition when reading out articles, she still could be her “rival”.
According to Gathercole, the labs have been in discussion with car manufacturers to see the possibilities that artificial voice could provide for commuters keen to listen to the news.
Looking to the future, Gathercole is confident that voice technology will continue to improve and allow for opportunities across devices. “I would imagine (within five years time) it would be possible to switch between audio and text and video across multiple devices,” he said. “So whichever one you happen to be near at the time it’ll carry on the story,” he said.