Google, Voiceitt each extending speech recognition capabilities

Google is working on new Google Search features designed to teach people foreign languages, The Information reports.

Reportedly dubbed Tivoli, the new project will be rolled out later this year, and will initially work via text, with the speech recognition capabilities potentially being added at a later time via Google Assistant.

Tivoli’s development started roughly two years ago using the neural conversation AI model at the time known as Meena (now LaMDA).

The new solution is designed to enable individuals to learn new languages and expand their earning potential by making them more employable.

LaMDA: an overview

The latest version of LaMDA was showcased at this year’s Google I/O conference.

Without any clear domain-specific training beforehand, LaMDA was able to talk from the point of view of two very disparate objects, respectively the planet Pluto and a paper airplane.

During the demo conversation, LaMDA was able to research and discuss concrete facts and events gathered in real-time from the internet, showcasing the potential of the new solution.

The model is based on a machine learning technique known as transformers, which is reportedly capable of understanding not only language but also emotions.

Voiceitt launches new speech recognition app

Accessible speech recognition firm Voiceitt has released an assistive technology app exclusively for Apple devices.

The Voiceitt app is temporarily free, to allow users to try it, and enables individuals to navigate their environments and control smart home devices using their own voice commands.

According to the company, the new solution aims to support people with speech and motor impairments, particularly during the pandemic.

“Everyone deserves to be able to express themselves and to be understood,” said Voiceitt CEO and Co-founder Danny Weissberg.

“With this launch, countless people with non-standard speech will be able to use their own voice to easily communicate with caregivers, loved ones, and even their smart home devices,” he added. “Voice recognition technology is finally becoming accessible to everyone.”

The Voiceitt app is also compatible with Alexa-powered devices and is based on the company’s machine learning and state-of-the-art speech analysis technologies.

Its algorithms can recognize not only atypical speech but also speech patterns including utterances, the cadence of speech, breathing pauses, and non-verbal sounds, in multiple languages.

Article Topics

AI  |  biometrics  |  Google  |  research and development  |  speech recognition  |  voice biometrics