Apple engineers are hard at work improving Siri, if three new posts on a company blog are any indication.
The three articles were posted to Apple’s Machine Learning Journal this month, and all three detail various ways that the company’s AI division is improving its proprietary digital assistant. The articles are definitely academic and information-dense, but they do hint at several ways that Apple is working to give Siri a boost.
Chief among the research is a new application of deep learning technology that improves Siri’s voice and speech patterns, explained in Vol. 1, Issue 4 of the journal. Apple first applied the tech to the digital assistant in iOS 10, and will continue improving the deep learning-based system in iOS 11. The end result is a Siri voice that’s smoother and more natural, and allows the assistant’s “personality to shine through.”
The article concerning this tech, helpfully, contains some voice samples that compare and contrast Siri’s speech patterns in iOS 9, iOS 10, and the upcoming iOS 11. In addition to the improved speech synthesis tech, Apple has apparently hired a new female voice actor for the iOS 11 version of Siri. “As a result, the new US English Siri voice sounds better than ever,” Apple wrote.
Additionally, the other two blog posts detail other methods of improving Siri. The first, Vol. 1 Issue 2, explains a new technique to train Siri to be better in “new languages and over new audio channels.” Issue 3 goes in-depth on an approach to improve the way Siri displays text-based information. Followers of Apple’s research will remember that the Machine Learning Journal’s first issue explained a new method of improving AI image recognition and computer vision.
A better-sounding voice is just one of several improvements slated to hit the digital assistant this fall. Other upgrades that iOS 11 will introduce include a slew of system-wide functionalities, such as an improved translation feature and the ability to fine-tune and customize Siri’s capabilities and sync those settings across a network of devices.
Apple first debuted its Machine Learning Journal in July as part of a larger effort to bring more transparency to the company’s AI and machine-learning research. Historically, Apple has kept much of its research secret — but changed its policy last year to allow for its team to publicly publish their research.