New in OS X: Get MacRumors Push Notifications on your Mac

Resubscribe Now Close

Apple's Latest Machine Learning Journal Entry Focuses on 'Hey Siri' Trigger Phrase

Apple's latest entry in its online Machine Learning Journal focuses on the personalization process that users partake in when activating "Hey Siri" features on iOS devices. Across all Apple products, "Hey Siri" invokes the company's AI assistant, and can be followed up by questions like "How is the weather?" or "Message Dad I'm on my way."

"Hey Siri" was introduced in iOS 8 on the iPhone 6, and at that time it could only be used while the iPhone was charging. Afterwards, the trigger phrase could be used at all times thanks to a low-power and always-on processor that fueled the iPhone and iPad's ability to continuously listen for "Hey Siri."

In the new Machine Learning Journal entry, Apple's Siri team breaks down its technical approach to the development of a "speaker recognition system." The team created deep neural networks and "set the stage for improvements" in future iterations of Siri, all motivated by the goal of creating "on-device personalization" for users.

Apple's team says that "Hey Siri" as a phrase was chosen because of its "natural" phrasing, and described three scenarios where unintended activations prove troubling for "Hey Siri" functionality. These include "when the primary users says a similar phrase," "when other users say "Hey Siri"," and "when other users say a similar phrase." According to the team, the last scenario is "the most annoying false activation of all."

To lessen these accidental activations of Siri, Apple leverages techniques from the field of speaker recognition. Importantly, the Siri team says that it is focused on "who is speaking" and less on "what was spoken."
The overall goal of speaker recognition (SR) is to ascertain the identity of a person using his or her voice. We are interested in “who is speaking,” as opposed to the problem of speech recognition, which aims to ascertain “what was spoken.” SR performed using a phrase known a priori, such as “Hey Siri,” is often referred to as text-dependent SR; otherwise, the problem is known as text-independent SR.
The journal entry then goes into how users enroll in a personalized "Hey Siri" process using explicit and implicit enrollment. Explicit begins the minute that users speak the trigger phrase a few times, but implicit is "created over a period of time" and made during "real-world situations."

The Siri team says that the remaining challenges faced by speaker recognition is figuring out how to get quality performance in reverberant (large room) and noisy (car) environments. You can check out the full Machine Learning Journal entry on "Hey Siri" right here.

Since it began last summer, Apple has shared numerous entries in its Machine Learning Journal about complex topics, which have already included "Hey Siri", face detection, and more. All past entries can be seen on

Top Rated Comments

(View all)

21 months ago
Apple can post all the journals and blog entries they want. Siri sucks. (When compared to others)

As I’ve said before, they need to fire the entire team and hire new. Poach if needed. They need to JOBS THIS SOB.
Rating: 8 Votes
21 months ago
Blah blah blah, neural networks, blah blah blah, set the stage for improvements, blah. They need to actually *make* those improvements and quickly. Either that or allow us to use Google assistant or Alexa natively on the iPhone. I'm tired of Siri telling me about web searches. It's hard to believe how Apple dropped the ball that they brought to the game.
Rating: 6 Votes
21 months ago

I thought they did say 'Computer' often before issuing commands in Star Trek? Maybe my memory is fuzzy though.

Computer was always the trigger word in Star Trek.
Rating: 3 Votes
21 months ago
They're whole siri system needs an overhaul.
Its lagging so far behind others, dictation is awful, through bluetooth is worse. its just all around poor.

Even spotify's new voice control is lightyears more accurate than siri :rolleyes:
Rating: 3 Votes
21 months ago
I had to disable this trigger. I can't say 'serious' anywhere in the vicinity of my iPad Pro.
Rating: 2 Votes
21 months ago
I can't say I've experienced too many unintended activations, but I do still wish apple would allow us to change the triggering phrase to whatever we wanted. That may help with unintended activations for each person.
Rating: 2 Votes
21 months ago
Siri will always be bad if all it will ever be is an "on-device" AI assistant. Without deep cloud integration it will never compete well.

Apple is trying to avoid the idea they need access to some of your information available on their cloud servers in order to make Siri a good product, so they are hyping up this idea that your phone becomes a highly specialized AI neural network, which sounds great for marketing, but in reality all it means is that Siri will find and access stuff on your phone with a massive amount of excessive engineering involved.

What Apple needs to do is recognize that there is a good balance between access and aggregation of your data vs having too much private data at their disposal, but I don't think Apple will win in this market if they draw a line land bury their heads in the sand and try to put all the power on your phone instead of taping into the power of the cloud.

I do not believe that an all or nothing approach Apple assumes about accessing consumer data is a wise business decision. Apple is hiding away from the real technical challenge of finding the right balance to make Siri a competitive assistant, so they are hyping up some nonsense AI neural network ******** that will be built into iPhone which will make is suck just as bad 5+ years from now if all it can do is scrape data on your phone and find a song or picture or web link for you.

This is once again a situation where an Apple CEO sticks foot in mouth and makes a huge public claim about how Apple will never do something, and that forces the companies hand for the next decade. Tim Cook went too hard trying to make Apple the darling of consumer privacy rights groups so now Apple will never be able to compete with Google or Amazon and try to trump up some BS iPhone AI technology platform claim trying to promote more iPhone sales in the future (and an increase in their prices as well), while Google and Amazon continue to advance their platforms by fully leveraging cloud integration.

This also isolates Apple from risk, such as what Facebook is challenged by now; by never TRYING, Apple never has to worry about failing, but its companies like Facebook that trail blazes and will ultimately strengthen the market by exposing critical failures and correcting them instead of companies like Apple that hide from innovation and wait until others have failed to figure out the minimum to move forward with.

Those who can't go into PR. Apple is the king of PR. I don't need massive amounts of AI to play a song on my iPhone, I want Siri to at least match capabilities of Google and Amazon, but it will never do so being an on-device assistant.
Rating: 2 Votes
21 months ago
How about they change the name "Siri" to something else?
Rating: 2 Votes
21 months ago
What irritates me most about Siri is the advertising that portrayed Siri as an intelligent, responsive, and efficient tool. Meanwhile, unless I’m mistaken, Siri was the first Apple product to be publicly released while still labeled a beta. In other words, Apple—who has a self-proclaimed standard of deeming half-baked tech unworthy—nevertheless let their marketing group exaggerate Siri’s readiness. Obviously, Apple didn’t apply the same scrutiny to their own invention as they do to outside ones.

Makes One wonder, what other product capabilities and benefits might Apple be exaggerating?
Rating: 1 Votes
21 months ago

Also, imagine walking into a room and having to know the name of the voice interface.

"Hey, Siri, turn on the lights." (Nothing happens.)

"Hey, Cortana, turn on the lights." (Still nothing.)


I walk into the room and turn a light switch on. Works pretty damn well ~99.99% of time ;-)
Rating: 1 Votes

[ Read All Comments ]