 |
| The Google Assistant will soon understand your voice better |
Google Assistant is the best digital assistant out there, thanks to the company's innovations in machine learning and the way it accesses all aspects of our lives, from web searches to smart home gadgets. . However, there is still room for improvement.
The company's goal is to make chatting with Assistant as easy and seamless as chatting with a friend or relative, but despite Assistant's regular upgrades, that goal remains elusive.
In March 2021, Google began using Unified Learning on Android to improve the accuracy of the important word Hey Google to activate its own voice assistant.
Unified learning is a machine learning approach that trains algorithms on multiple distributed machines or servers containing local samples of data without exchanging it.
It now appears that the upcoming personal voice recognition feature will help the Google Assistant better recognize repeated words and names.
According to Google's latest version of the Android app icon, personal speech recognition appears in the Assistant settings.
The company describes the feature: Recordings about that device are saved so the assistant can better understand what you're saying. Recordings are stored via this device and can be deleted at any time by disabling personal voice recognition. You can get more information.
For more information, link to a recent support article describing the company's use of federated learning to improve keyword activation using audio recordings stored on a user's device to improve templates like Hey Google Discovery.
The assistant learns model tuning from audio data. It also sends a summary of the form changes to the company's servers. These summaries are aggregated across multiple users to provide a better form for everyone.
Google Assistant understands your voice better
The upcoming feature aims to bring these machine learning-based improvements to real-world processor commands. They are particularly associated with common nouns and words.
Recordings of past speeches are stored on the device and analyzed to make future versions more accurate.
On devices like the second-generation Nest Hub and Nest Mini, Google uses a machine learning chip to process recurring requests locally, speeding response times. This concept can now be extended beyond smart home devices to Android.
And given Google's stance on the Assistant and voice privacy, it's likely a subscription feature that helps improve the Assistant.
According to the company's available feature description, the recordings remain on the device and are deleted when the feature is disabled.
Meanwhile, Google says it's turning off personal speech recognition: If you turn the feature off, the Assistant will be less accurate at recognizing names and other words you say frequently. All tones used to improve speech recognition will be removed from the device.
It is unclear when this feature will be introduced or to what extent it will be improved. But Google showed at I/O 2022 how ancillary conversations could become more natural in the coming year. The assistant has to largely ignore interruptions, graceful pauses, and other self-corrections.
It seems that Google wants its assistant to better understand the commands and words that are most specific to you with the new feature.