Best printers for Home

Best Mixer Grinder

Best wired Earphones

Best 43 Inch TV in India

Best Wi Fi Routers

Best Home Theatre in India

Smart Watch under 5000

Best Laptops for Education

Best Laptop for Students

Researchers at <keyword id="10387" type="CITY" weightage="100" keywordseo="Microsoft">Microsoft</keyword> have released details of new speech recognition technology that they say translates conversational speech as well as a human. "We've achieved human parity," says Microsoft's chief speech researcher <keyword id="3764665" type="UNKNOWN" weightage="20" keywordseo="xuedong-huang">Xuedong Huang</keyword> in an announcement. "This is an historic achievement."<br />
<br />
For a considerable length of time science fiction movies have been envisioning the future as one where humans talk to machines pretty much like family and friends. In reality, be that as it may, using voice to interact with machines has been maddeningly frustrating, with Siri frequently mixing up "open up my email" for "look into some fail," for instance. This is evolving.<br />
<br /> A group at Microsoft's <keyword id="107950" type="TECHNOLOGY" weightage="20" keywordseo="artificial-intelligence">Artificial Intelligence</keyword> and Research group have built a <keyword id="107996" type="TECHNOLOGY" weightage="20" keywordseo="voice-recognition">voice recognition</keyword> system that commits errors at an indistinguishable low rate when compared to professional human transcribers. The researchers investigated an assortment of recurrent neural networks and had the most success with a LSTM (Long Short-Term Memory) model. The model was trained with a 30k-word vocabulary for 2,000 hours on a server farm comprising of multiple GPUs.<br />
<br /> Microsoft plans to utilize the technology in <keyword id="1865287" type="UNKNOWN" weightage="60" keywordseo="cortana">Cortana</keyword>, its personal voice assistant for Windows and the Xbox One, and also speech-to-text transcription software<br />
<br />
The group's next objective is to enhance the engine's robustness so it can be used as a part of real-life situations, for example, on crowded city streets or while driving. They additionally would like to inevitably motivate it to work with multiple users simultaneously.

Microsoft just released the
future of Voice Recognition

<p><div class="KonaFilter image-container display-table float_right"><div><div class="image on-image"><p class="source"><span>Microsoft</span></p><p class="caption">Microsoft Research Technical Fellow Xuedong Huang (left) and his team reached a major breakthrough in speech recognition.</p></div><p></p></div></div>In October 2016, in a big milestone for artificial intelligence, <a href="http://www.businessinsider.com/microsoft-beats-humans-at-speech-recognition-2016-10">Microsoft unveiled a system </a><a href="http://www.businessinsider.com/microsoft-beats-humans-at-speech-recognition-2016-10">that can transcribe the contents of a phone call as well or better than human professionals</a>.</p><p>But while Microsoft's system had fewer transcription errors than the average human transcriptionist, it still couldn't best a team of trained humans. So, the world of academia fired back with a new challenge: Lower the error rate to below what human teams can do. </p><p>Now Microsoft has done just that. In a <a href="https://www.microsoft.com/en-us/research/blog/microsoft-researchers-achieve-new-conversational-speech-recognition-milestone/">blog entry on Sunday</a>, Xuedong Huang, Microsoft Research's chief speech scientist, reported that the company had broken even that barrier.</p><p>It's a major milestone, Huang wrote. And it gives the company a sound foundation to go from mere transcription to understanding the meaning of what's being said, he said. Speech recognition is a fundamental building block for building more robust artificial intelligence.</p><p>"Moving from recognizing to understanding speech is the next major frontier for speech technology," Huang wrote.</p><p>Microsoft's voice recognition system has been improving rapidly. Transcription accuracy is judged by error rates; i.e., the portion of words a system gets wrong out of a given recording of speech. That error rate is determined using Switchboard, a standard test for voice transcription accuracy widely used in the industry, including by IBM and Google.</p><p>As recently as September 2016, Microsoft's error rate, according to Switchboard, was 6.3%,<span> which means that out of every 100 words the system was getting more than 6 wrong</span>. By comparison, a single human transcriptionist has an average error rate of 5.9%, and a team of trained humans clocks in with an error rate of around 5.1%.</p><p>Microsoft matched the former error rate in October and just beat the latter.</p><p>That's far sooner than the company expected. Indeed, back in 2015, Huang himself told Business Insider that building a system capable of surpassing a human at transcription was "four to five years away." Less than two years later, we're well past that point.</p><p>Still, challenges remain. Microsoft's transcription system is patterned after the audio coming from a nice, stable landline telephone, <a href="http://www.businessinsider.com/microsoft-beats-humans-at-speech-recognition-2016-10">Geoffrey Zweig, formerly a principal researcher at the company, told Business Insider</a> last October. The next frontier for voice recognition is to accurately transcribe speech even when it's coming over a lousy cell connection or an echoing McDonalds drive-thru speaker.</p><p>Speech science "still has many challenges to address, such as achieving human levels of recognition in noisy environments with distant microphones, in recognizing accented speech, or speaking styles and languages for which only limited training data is available," Huang wrote in his blog post on Sunday.</p><p class="tagline">Get the latest Microsoft stock price<a href="http://markets.businessinsider.com/stock/MSFT-Quote"> here.</a></p>    <div class="margin-top">        <h3>NOW WATCH: <a href="http://www.businessinsider.com/6-things-bill-gates-accurately-predicted-back-in-1999-microsoft-tech-book-business-speed-thought-2017-7">6 things in tech today that Bill Gates accurately predicted back in 1999</a></h3>        <div>                    </div>    </div>

Microsoft's voice-recognition tech is now better than even teams of humans at transcribing conversations

<p><div class="KonaFilter image-container display-table float_right"><div><div class="image on-image"><p class="source"><span>Dan DeLong</span></p><p class="caption">The Microsoft Research team responsible for setting the new milestone. Back row, left to right: Wayne Xiong, Geoffrey Zweig, Frank Seide. Front row, Xuedong Huang, Dong Yu, Mike Seltzer, Jasha Droppo and Andreas Stolcke.</p></div><p></p></div></div>In December 2015, <a href="http://www.businessinsider.com/microsoft-chief-scientist-xuedong-huang-on-the-future-of-speech-recognition-2015-12">Microsoft Chief Scientist of Speech Xuedong Huang told Business Insider</a><span> that <span>"in the next four to five years, computers will be as good as humans" at understanding the words that come out of your mouth.</span></span></p><p><span><span>Less than a year later, and Microsoft just set a record with the announcement of a system that can transcribe the contents of a phone call with "the same or fewer errors" than real actual human professionals trained in transcription. </span></span></p><p><span><span>It's a huge milestone for speech recognition, even as gadgets like <a href="http://www.businessinsider.com/apple-iphone-7-siri-airpods-amazon-alexa-echo-2016-9">Amazon Echo and Apple's Airpods</a> prove that voice is going to play a big role in the future of technology.  And by Huang's standard, that's mission accomplished.</span></span></p><p><span><span>"We were able to move more quickly than we anticipated" thanks to advancements in artificial intelligence and acoustic technology, Microsoft Principal Researcher Geoffrey Zweig tells Business Insider, and "we were able to get here faster."</span></span></p><h2><span><span>Switchboard test</span></span></h2><p><span><span>Back in the 1990's, the National Institute of Standards and Technology (NIST) released a whole bunch of recorded phone conversations in English, Spanish, and Mandarin, called "<a href="http://www.itl.nist.gov/iad/mig/tests/ctr/2000/h5_2000_v1.3.html">Switchboard</a>," as a way to keep things fair for the field of speech recognition research. Everybody is working from the same data, so nobody can cheat.</span></span></p><p><span><span>Since then, lots of companies, including IBM, Google, and Microsoft itself, have used the Switchboard test as one of the main ways to check the accuracy of their speech recognition software. </span></span></p><p><span><span>A phone call is a great test because, as in real life, people mumble, mutter, cough, and otherwise stumble over their words, making automatic transcription a "much more difficult task" than it would be under laboratory conditions, Zweig says.</span></span></p><p><span><span><div class="KonaFilter image-container display-table"><div><div class="image on-image"><p class="source"><span>Microsoft</span></p><p class="caption">Microsoft Distinguished Engineer Xuedong Huang</p></div><p></p></div></div></span></span></p><p><span><span></span></span>Back in September, Huang announced via blog entry that Microsoft Research had achieved an error rate on the Switchboard test of 6.3%. He said Microsoft's error rate was believed to be the best in the whole industry, and only a hair above the 5.9% average error rate among professional transcribers.</p><p>So, Microsoft made some tweaks to the model, and did what Zweig says nobody had ever done before: Took the Switchboard test and gave it to those professionals to transcribe, to compare the results.</p><p>Why had nobody taken that step before? Maybe because it was "beyond the imagination" that even the best systems were even close to matching a human, Zweig speculates. Regardless, the results came back and NIST verified them.</p><p>Microsoft had officially built a speech recognition system that was better than a human.</p><h2>What's next?</h2><p>In the shorter term, this technology is going to make Microsoft's Cortana virtual assistant much better at understanding you. In the long term, Zweig says, Microsoft is working hard at using this successful model and then tweaking it for more situations.</p><p>Right now, it's optmized for listening in on a conversation on a nice, stable landline telephone. With the core speech recognition algorithms all stable, now they can tweak it to better understand you when you're on a noisy city street, or an echo-y conference room, or even using a McDonalds drive-thru.</p><p>And the more people use it in all these situations, the better it gets for everyone, Zweig says, as the algorithms learn and improve.</p><p>"This is a technology that's constantly improving," Zweig says.</p><p><div class="KonaFilter image-container display-table"><div><div class="image on-image"><p class="source"><span>YouTube/Microsoft</span></p><p class="caption">Microsoft Cortana on Android</p></div><p></p></div></div></p><p>And in general, this <keyword id="41551" type="ENTERTAINMENT" weightage="60" keywordseo="science">science</keyword> is a huge and important step forward as speech recognition becomes ever more important to the future of technology. With the ability to understand the words coming out of your mouth, it's a <a href="http://www.businessinsider.com/googles-newest-software-is-named-parsey-mcparseface-no-seriously-2016-5">solid foundation on which to build better, smarter artificial intelligence that can find the context around the words</a>.</p><p>"We've actually managed to advance the technology of speech recognition," says Zweig.</p>    <div class="margin-top">        <h3>NOW WATCH: <a href="http://www.businessinsider.com/hand-talk-app-translate-speech-sign-language-hugo-brazil-libras-2016-9">This virtual character can translate speech into sign language</a></h3>        <div>            <div>        <div class="ooyala-video-player" data-video-autoplay="false" data-video-id="1yMHByNTE6mR7QPtP6V0vXnG0QI5vUNX" data-player-variable="ooyalapopvideo" id="ooyalaplayer_popvideo" style="width:1280px;height:720px"></div>        </div>        </div>            </div>

Microsoft built technology that's better than a human at understanding a conversation

<p></p><div class="KonaFilter image-container display-table float_right"><div><div class="image on-image"></div><p></p></div></div>As frustrated as you might get that voice-controlled tools like Apple Siri and Microsoft Cortana don't always understand you, it used to be a lot worse.<p><a href="http://www.businessinsider.com/where-is-google-taking-us-2015-7">Earlier this year</a>, Google announced it had gotten its speech recognition error rate down to 8%.&#xa0;</p><p>But Microsoft Distinguihsed Engineer and Chief Scientist of Speech Xuedong Huang says that it's a vast improvement.</p><p>When Microsoft made its first-ever speech recognition technology available alongside Windows 95, a project Huang headed up, the error rate was "almost 100%," he says.</p><p>If you chart it out, Huang says, that means that on average, speech recognition has gotten 20% better every single year for the last twenty years. Which means that the end is in sight.</p><p>"In the next four to five years, computers will be as good as humans" at understanding the words that come out of your mouth, Huang says.</p><p>But for Huang, Microsoft, and the tech world in general, the end of this road is the beginning of the next phase: building real artificial intelligence.&#xa0;</p><h2>"Invisible revolution"</h2><p>With "total parity" between human and computer&#xa0;understanding on the visible horizon, Huang says, it means that the world of speech science has a firmer foundation on which to work on giving computers&#xa0;actual artificial intelligence.</p><p>"To understand a word is easier than understanding the context," Huang says.&#xa0;</p><p>But with tools like Microsoft Cortana, Google Now, Apple Siri, and Amazon Alexa, we have consumer-facing apps that are slowly but surely getting better at figuring out not only what you said&#xa0;- but also what you meant. It means that you can start to have more complex conversations with your gadgets.</p><p>This means that we're on the cusp of an "invisible revolution," Huang says, as speech becomes an accepted and useful interface for computers, and artificial intelligence becomes a reality.</p><p>It's something that's been a long time coming for Microsoft. Witness&#xa0;Bill Gates demonstrate the Microsoft MiPad (seriously), a device running prototype voice recognition software created by Huang's team, at the 2001 Consumer Electronics Show:</p><p><div></div></p><p>The MiPad never came to market. But the world of speech technology marches on.</p><h2>The world of tomorrow</h2><p class="embed-spacer">Huang is heavily involved with <a href="http://www.businessinsider.com/microsoft-project-oxford-can-see-when-youre-happy-2015-11">Project Oxford</a>, the company's set of machine learning tools for image and speech recognition. If you used fun Microsoft sites like <a href="http://www.businessinsider.com/microsoft-built-a-site-to-guess-how-old-you-are-2015-4">How-Old.net</a> or <a href="http://www.businessinsider.com/microsoft-rates-your-mustache-2015-11">MyMoustache</a>, you've gotten some hands-on experience with what it can do.&#xa0;</p><p>Project Oxford is available to developers everywhere, meaning developers can put the&#xa0;technology in their own apps.&#xa0;</p><p>And just as Microsoft Cortana can listen to your spoken questions and give smart answers, Project Oxford lets developers of consumer apps, business software, and everything in between build&#xa0;technology with whom&#xa0;you can hold a decent conversation.&#xa0;</p><p></p><div class="KonaFilter image-container display-table"><div><div class="image on-image"></div><p></p></div></div><p>It means the rise of one interface&#xa0;- speech&#xa0;- that can control every kind of device, anywhere in a home. And with Microsoft Project Oxford, and the Microsoft Azure cloud that underpins it, Huang says Microsoft is in a great position to be at the center of the revolution.</p><p>"It took us 20 years to reach that goal," Huang says.</p><p>Even as Microsoft works on AI, the company has already started to look ahead to what's next, Huang says.</p><p>Indeed, he says that the <a href="http://www.businessinsider.com/why-microsoft-xbox-kinect-didnt-take-off-2015-9">Xbox Kinect sensor</a>, which let people control Xbox video games with voice and motion, was actually born of Microsoft Research's first cracks at building systems that could understand both speech and gestures to discern meaning.&#xa0;</p><p>Eventually, this will just be a new normal, Huang believes. Kids will grow up with these kinds of artificial intelligence systems, and they will be taken for granted &#xa0;as standard methods for&#xa0;interacting with technology.</p><p>"We are creating a new generation," Huang says.</p>    <div class="margin-top">        <h3>NOW WATCH: <a href="http://www.businessinsider.com/weird-siri-responses-easter-eggs-iphone-2015-7">11 things you can ask Siri to get the most bizarre and hilarious answers</a></h3>        <div>            <div>        <div class="ooyala-video-player" data-video-autoplay="false" data-video-id="pvbmowdjryYQWUz0YgYfFii1E_gjvKxT" data-player-variable="ooyalapopvideo" id="ooyalaplayer_popvideo" style="width:1280px;height:720px"></div>        </div>        </div>            </div>

A chief scientist at Microsoft says we're less than five years away from computers understanding us perfectly

<p><div class="KonaFilter image-container display-table float_right"><div><div class="image on-image"><p class="source"><span>Microsoft</span></p><p class="caption">Microsoft Corporate VP of Microsoft Research Peter Lee</p></div><p></p></div></div>At today's <a href="http://www.structuredata.com">Structure Data</a> event in San Francisco, Microsoft Research head Peter Lee told the story of a public speaking nightmare that highlights just how difficult it is to build machines that can completely understand human language.</p><p>Some time in 2015, Lee says he was giving a keynote presentation on the power of technology and the potential of artificial intelligence at a conference in San Francisco. </p><p>To drive his point home, Lee was having his presentation translated into Mandarin over the stage speakers in real-time via Skype Translate.</p><p>Skype Translate, a project born out of Microsoft Research's speech laboratories, was <a href="http://www.businessinsider.com/microsoft-research-under-pressure-from-satya-nadella-2016-1">rushed through development by CEO Satya Nadella</a> to show off how <a href="http://www.businessinsider.com/microsoft-research-future-visions-2015-12">Microsoft's bleeding-edge science</a> can be used in real products.</p><p>Part of Lee's speech involved a personal story of growing up in a "snowy town" in upper Michigan. He noticed that most of the crowd was enraptured - except for a few native Chinese speakers in the crowd who couldn't stop giggling.</p><p>After the presentation, Lee says he asked one of those Chinese speakers the reason for the laughter. It turns out that "snowy town" translates into "Snow White's Town."</p><p>Which seems innocent enough, except that it turns out that "Snow White's town" is actually Chinese slang for "a town where a prostitute lives," Lee says. </p><p>Whoops.</p><p>Lee says it wasn't caught in the profanity filters because there weren't actually any bad words in the phrase. But it's the kind of regional flavor where a direct translation of the words can't bring across the meaning. </p><p>In fact, Lee says that this kind of nuance is something that's really difficult for the Skype Translate team. Another good example is the "umms" and "ahhs" that you may include in your speech - Skype Translate is totally capable of wiping them out in translations, but what if those pauses and stops are part of how you're conveying meaning?</p><p class="p1">"Maybe those 'ums' and 'uhs' are part of your charm," Lee says.<div class="KonaFilter image-container display-table"><div><div class="image on-image"><p class="source"><span>Skype</span></p><p class="caption">Skype Translate</p></div><p></p></div></div></p><p class="p1">Still, there's cause for optimism that this problem will be solved. In the next four to five years, computers will be as good as humans" at understanding the words that come out of your mouth," <a href="http://www.businessinsider.com/microsoft-chief-scientist-xuedong-huang-on-the-future-of-speech-recognition-2015-12">Microsoft Research chief scientist <span>Xuedong</span> Huang told Business Insider</a> in late 2015.</p><p class="p1">Other tidbits from Lee's talk: Microsoft Research is placing the bulk of its efforts behind "<a href="http://www.businessinsider.com/microsoft-hybrid-quantum-computer-2015-10">quantum computing</a>," a field <a href="http://www.businessinsider.com/googles-d-wave-2x-quantum-computer-2015-12">understood by scientists</a> to be possible of creating computers that are hundreds more times powerful than we have today.</p><p class="p1">In fact, Lee says that quantum computing right now is like speech recognition was for most of the 1990s: Lots of good work being done, but waiting for the next big breakthrough before it's ready for prime time.</p><p class="p1">"The way I tried to explain it to [Microsoft CEO] Satya Nadella, we worked on [speech recognition] for almost a decade with no practical discernible improvement, and then we hit this tipping point," Lee says.</p><p class="p1"> </p>    <div class="margin-top">        <h3>NOW WATCH: <a href="http://www.businessinsider.com/man-skypes-parents-while-skydiving-2015-12">A guy Skyped his parents while jumping out of a plane</a></h3>        <div>            <div>        <div class="ooyala-video-player" data-video-autoplay="false" data-video-id="0xYXJseTqo3lcRGecz9CKhkvg5jGq5Zu" data-player-variable="ooyalapopvideo" id="ooyalaplayer_popvideo" style="width:1280px;height:720px"></div>        </div>        </div>            </div>

You Searched For "xuedong huang" and got 5 results

Microsoft just released the future of Voice Recognition

Gaurav Sangwani

Microsoft's voice-recognition tech is now better than even teams of humans at transcribing conversations

Matt Weinberger

Microsoft built technology that's better than a human at understanding a conversation

Matt Weinberger

A chief scientist at Microsoft says we're less than five years away from computers understanding us perfectly

Matt Weinberger

This Microsoft exec's hilarious presentation fail shows why computer translation is so difficult

Matt Weinberger

Popular Keywords

Buying Guides