NEW FROM "BY DO"

Baidu’s new smartphone keyboard lets you type sentences with just one tap (and a murmur)

This article was written using Baidu’s voice recognition keyboard—the raw transcript is under each paragraph, designated by pink type. The final version has also been only very lightly edited in order to most accurately show the efficacy of Baidu’s software.

Earlier this year, Google CEO Sundar Pichai said that 20% of Google searches on mobile phones are now entered by voice. It’s a huge metric that shows we’re getting more and more comfortable talking to our silicon sidekicks, and that voice recognition is getting to a point where it can actually be useful.

Earlier this year, google ceo sooner pichai said that twenty percent of google searches on the mobile phones are now done by voice. it’s a huge metric that shows we are getting more and more comfortable talking to our silicon side kicks, and that voice recognition is getting to a point where I can actually be useful.

Researchers at Baidu, the Google of China, are pushing this idea forward with a smartphone keyboard centered around voice, TalkType. The keyboard has a few other commonplace functions, like quickly summoning GIFs and sharing nearby eateries via Yelp, but the main star is voice recognition. As demonstrated in an earlier project with Stanford University, Baidu thinks that the future of smartphone input lies in speech: humans will talk to their phones, and not tap on glass, to enter their information.

Researchers at bido, the google of china, are pushing this idea forward with a smartphone keyboard at centered around voice, tag type. The keyboard has a few other commonplace functions, like quickly summoning gifts and sharing nearby eateries via yelp, but the main star voice recognition. As demonstrated in an earlier project with stanford by do thinks that the future of inputs for phones lies in speech comm humans will talk to their phones not tap on glass to enter their information in the future.

All the trimmings.
All the trimmings. (Quartz/ Dave Gershgorn)

After trying the app, it’s clear that software has a ways to go. Dictating punctuation is still an arduous task, and it’s inefficient to speak to the app like it’s a human being. Rare words like company names or surnames are difficult for the keyboard to understand, a result of the kind of artificial intelligence used for the voice recognition. If the AI hasn’t heard previous of examples of a word, it’s extremely unlikely that the keyboard will figure it out. For instance, my last name Gershgorn, is not recognized. On my first attempt, the keyboard recognized “gersh corn,” and on the second attempt it recognized “gershon.” After I imported my contacts, which includes my own name, it found “gersh gorn,” “gersh garn,” and “gersh corn.” The keyboard still could not recognize my name but got most of the letters right.

Trying the app, it’s clear that software has a ways to go. Dictating punctuation is still an arduous task and there is no way to speak to the app like it’s a human being. Rare words like company names or surnames are difficult key word to understand, a result of the kind of artificial intelligence used to train the voice recognition. For instance, I last name , gershon, is not recognized. O me attempt, the recognized gersh corn, and on the second attempt recognized gershon. And when I imported the contacts , and said gersh gorn, gersh garn , gersh gorn , gersh corn , gersh gorn, the keyboard still could not recognize my name but got most of the letters right.

Speaking into your phone is still awkward, especially in public. I’m dictating this into my phone in a private phone booth, to avoid the angry glares of coworkers. Longer dictations like this one are also limited by how long the screen of the phone will stay activated without a user touching it.

Speaking into your phone is still awkward especially in public. I’m dictating this into my phone in a private phone booth avoid the angry glares of coworkers. Longer dictations, like this one are also limited by the active screen time of the phone com meaning how long the screen will stay activated without a user touching it.

Much like similar voice recognition software (like that offered by Google), the audio files, transcripts, and logs of your speech will be catalogued and sent to Baidu to help train the artificial intelligence software. By showing their software what worked and what did not, the company is able to train more precise voice recognition in the future.

Much like similar voice recognition software offered by google, the audio files transcripts and logs of your speech will be catalogued and sent to by de help train the artificial intelligence software. By showing their software what worked and what did not the company is able to train more precise voice recognition in the future.

The TalkType keyboard is only available for Android phones.

The talk type keyboard is only available on android phones for now, backspace. backspace delete

home our picks popular latest obsessions search