Apple has launched a speech study, designed to improve Siri’s voice recognition and acoustic models. The company states that they aren’t using the app as a marketing tool but rather collecting feedback from consumers about Siri’s performance.
According to TechCrunch, this study is worddocx only open to those invited by Apple and not all participants will be required to share any information with the company. The app itself is straightforward, providing a link to Apple’s license agreement and prompting participants to enter their ID number in order to get started.
This study employs federated learning, a privacy-preserving machine learning method. This process allows the company to train multiple hdxwallpaper copies of a speaker recognition model on each user’s device and send only those updated models to a central server. This permits them to utilize raw audio from users’ Siri requests on their own devices without disclosing it to third parties.
Furthermore, this allows the company telesup to monitor how well its assistant understands each user’s language and dialect. Thus, they can continuously develop and enhance their capabilities.
A key aspect of this improvement is a new technique for producing natural-sounding voices that can be used in Siri. This technology, called happn Neural text to speech (TTS), replaces acting talent-based recordings with audio that has been fed through an artificial neural network to generate spoken words. This marks an immense advancement for AI as it produces voices that sound much more authentic and less robotic.
Siri can now make better use of its on-device voice learning and acoustic modeling to boost its voice accuracy over time. This includes roobytalk tweaks like intonation, inflection and phrasing that sound more natural and human.
Apple has been steadily adding voice models to Siri over the past year, including an Irish Siri that can speak both Irish and English, as well as an Italian Siri that speaks both Italian and German.
Another change introduced in this beta version is a voice selection option for users. Siri currently defaults to male voices in some languages, but with these new options users will have the freedom to pick what kind of voice they prefer. This step helps combat bias that unfortunately still exists among voice selection processes on these types of apps.
Siri’s new voices offer a welcome diversity of options, enabling users to pick one that best suits their preferences and requirements. This marks an enormous victory for Apple and those seeking the ideal voice – finally they have someone who fits perfectly!