Friday, August 26, 2016

Siri's Voice to Benefit from Machine Learning Improvements in Apple's iOS 10

Siri

Extensive Improvement in Digital Personal Assistant – Siri


When iOS 10 had been introduced earlier in the year by Apple, one of the tent-pole additions in the software of the company had been extensive improvement in the digital personal assistant – Siri. Over the years, Siri has been progressing due to a focus on machine learning and new features combined in the mobile operating system, enabling third party application developers in gaining access on stock apps as well as their features. Siri would be getting a big enhancement with iOS 10 later in the year.

Siri’s voice could be more natural when iOS 10 tends to arrive this fall probably this September owing to its machine learning technology which Apple would be implementing according to an interview disclosed recently. Steven Levy had an impressive interview on Backchannel with various executives of Apple including Senior Vice President of Internet Software and Services, Eddy Cue, Phil Schiller and Craig Federighi.

Plenty of focus has been drawn on Siri, with improvements and how it would impact Apple’s progress. The interview did not begin with the statement with regards to Siri but expressed as to why Siri was so important for Apple’s progress and how would the digital personal assistant become over the years as well as why iOS 10 would be a big upgrading over all that has already been upgraded over a period of time.

Replacing Earlier Licensed Implementation


With the release of iOS 10, this fall, Siri’s voice would be the last of the four components which would be transformed by machine learning and a deep neural network is said to be replacing the earlier licensed implementation. Tom Gruber, head of advanced development for Siri, observed that when the responses of the assistant are yet being darned together from the central database of recordings, the machine learning tends to smooth out sentences, making Siri sound much more human.

The robotic- sounding voice of Siri has frequently been deceived and disapproved. Recently singer Barbra Streisand had contacted the CEO of Apple, Tim Cook, complaining regarding the way Siri pronounced her last name. Cook had replied that he would look into the problem in the forthcoming update which could have been a situation to the upgrade of Siri. The interview featuring executives such as senior VPs Eddy Cue together with Craig Federighi, observed that Apple had moved the voice recognition of Siri to a neural net-based system.

Enhanced Ability of Siri in Comprehending Commands


It is said that the technology has extremely enhanced the ability of Siri in comprehending commands. Federighi has commented that Apple had plenty of people working on the machine learning technology comprising of not only Siri but also things such as palm rejection for the Apple Pencil, though there is no central machine learning group.

Apple had moved Siri voice recognition to a neural-net based system for US users late one July day which went worldwide on August 15, 2014. Few of the earlier techniques seemed to stay operative, which comprises of the `hidden Markov model, though presently the system leverages machine learning techniques, together with deep neural networks, convolutional neural networks, long short-term memory units, gated recurrent unites and n grams.

Siri will still seem to look the same when users tend to upgrade, though now it has been supercharged with deep learning.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.