The Windows weakness no one mentions: Speech recognition - salazarlaure1957

This news report was originally publicised Oct. 7, 2016, and updated Crataegus oxycantha 10, 2017 with new information.

Windows has a feature it doesn't like-minded to talk about. Spell the OS lets you scribble notes with a stylus, log in with you grimace (or secure the Web) via Windows Howdy, and even order Cortana to set a reminder, what it's not so eager for you to do, apparently, is use its speech recognition engine to issue commands OR take voice bid.

The reason for its silence may cristal back 10 geezerhood, to when Microsoft product manager Shanen Boettcher demonstrated phonation dictation  inside Windows Vista—and flubbed it. The technology unbroken a low profile subsequently that, and nowadays, few users know you can dictate a document inside Windows.

If there were always a time for Windows to hear again, though, information technology would seem to be nowadays, when advances in computers and artificial tidings supply a overmuch better foundation for the engineering science. And it has.

At its Build 2017 developer group discussion, Microsoft launched a new Video Indexer preview that non only transcribes the video, but as wel identifies the speaker, provides optional translations in upfield to ball club languages, automatically generates subtitles, and guesses what objects operating theater overlays are on the screen. It even performs basic sentiment analysis, determining whether the words used are sensationalism OR negative. And IT's all searchable via a Web portal: If you only want to view the text from a specific speaker, you can.

Cross Hachman / IDG

Microsoft's Video Indexer pulls an incredible amount of information from video uploaded to it.

TV Indexer is an case of how Microsoft is applying AI to daily tasks. For example, the company showed off a PowerPoint Translating program procedure that will leave users to auto-configure a PowerPoint presentation in their autochthonic language. Video Indexer, though, goes far beyond.

According to the product handler for Video recording Indexer, Milano Gada, the indexer fundament't immediately discover every speaker in a video. But if a substance abuser identifies an "unknown" talker with their name, the entire database leave comprise updated with the slump data, he said. Video Indexer also quickly allows a video to be searchable, allowing consumers to skip right to where they'Re most interested.

Mark Hachman / IDG

Video Indexer justified auto-translates audio embedded in the TV.

That all begs the question: if Microsoft can surrender a solution like this for enterprise customers, wherefore can't it at least tap into the power of Cortana to deliver the same features for consumers?

Microsoft's secretiveness on words bidding

"This is such a great question," said Harry Shum, the enforcement frailty president overseeing Microsoft's language-recognition research, atomic number 3 well as Cortana and Bing, when asked hold up year about dictation's ulterior within Microsoft Office. "In that respect is really no reason out why IT is non playing a much more prominent role yet."

We decided to give it another chance: We delved into Windows' sound command features to see how they compared to more new speech-based technologies.

Deman Give voice 2016 about dictation, and it's like the app has never true detected the full term. Password displays a similar reply for "lecture recognition."

Why words recognition crapper't be too perfect

Some of us still think about vocalization dictation in the same wayDoonesbury lampooned the Apple Newton, turn "I am writing a test sentence" into "Siam fighting microscopic picke." And you'd be forgiven for thinking so, as well: Windows Speech Recognition is powered by the Microsoft Lecture Recognizer 8.0, which has remained literally unchanged since Vista. Shum known as it a "grandpa" technology.

Whathas changed, however, is the computer hardware: Hearing for and interpreting speech requires far less processing mogul than a decade ago. The select of integrated set out mics within PCs look-alike the Surface Book mean that holy headsets aren't necessarily required to achieve superior accuracy. Voice dictation for the masses is present, right?

When I tested Windows' speech capabilities, however, I experienced firsthand the tigerish perfection that's required for the organisation to follow usable. This story has 1,028 words in it, including subheadings. If you ill-used part dictation software to write it, a 95.0% truth rate would have in mind you'd have to correct much fifty dollar bill mistakes. That gets old swift.

In my tests, supported a methodological analysis I developed for another speech recognition product I'm testing, Windows produced an accuracy rate of 93.6%, That's pretty bad in theory, and somewhat can the holy package I'm trying. Windows as wel had an unexhausted habit of interjecting the word "comma butterfly" when I was dictating the punctuation mark mark. The speech community seems split on whether comparatively minor mistakes like-minded this are evidentiary.

That, naturally, was just the baseline. As anyone who's used dictation software program can tell you, the key to accuracy is training. Over clip, a voice dictation program learns your accent, whether you pronounce the "a" in apricot suchlike "bad" or "emulator," and how to strain our unconscious verbal tics. I've seen Microsoft employees claim that, properly trained, Windows' speech recognition was 99% exact. Ten mistakes Oregon indeed per 1,000 row isn't bad at altogether.

Very few of us, though, probably want to pass the prison term breeding the software. Windows Speech Credit requires up to 10 proceedings to run through a few practice sentences, and IT feels like a lifetime. Cortana and Siri Don River't require any of the very setup time, as they've already been trained on millions of voice samples. There's something to be said for moment gratification.

Training speech inside Windows is a lengthy outgrowth. The frame-up time associated with Nuance's Dragon software is far shorter, perhaps a minute roughly. But current digital assistants recognize your words at once.

What makes Cortana (which you can use on your PC or call up) much better than Windows' own ancient voice dictation systems is her link to the massive computational ability of the Microsoft cloud. Microsoft can crunch and related to your voice input together with whatever other information Microsoft knows about you, generating the intelligence that is the individual of Cortana.

Microsoft dialogue up delivery recognition

Given Cortana's proven skills, you'd think speech would rent essence stage. Simply at Build 2016, executives said dictation capabilities won't constitute added to Office. Last October, though, chief executive Satya Nadella's keynote plow at its Ignite conference multi-color speech recognition as a critical constituent of Microsoft's future.

Take Skype Translator, for example. Microsoft's Star Trek-like universal transcriber depends upon three different strands of explore, accordant to Nadella: speech recognition, delivery synthesis, and machine displacement.

"Straight-grained inside of Word or Mind-set when you'Re writing a document we now don't have simple thesaurus-founded spell correction," Nadella said, adding that Office give the sack now even compensate for dyslexia. "We have complete computational lingual understanding of what you're building. Or what you're writing."

But not what you're expression, apparently.

Microsoft

Microsoft chief executive Satya Nadella stands next to NFL headliner Deion Sanders at Microsoft's Ignite group discussion. Has Microsoft fumbled its dictation opportunity?

During the same speech, Nadella bragged that Microsoft's speech algorithms achieved a word error rate of 6.9 percent exploitation the NIST Switchboard test. That sounds disobedient: that's truth of near 93.1 percent. But the Patchboard screen uses sample rates of just 8KHz, almost the prize of a telephony conversation in the year 2000. Windows Media Sound 10, the codec within OneNote, can capture audio at up to 48KHz, providing much more accurate samples.

I suppose it's pretty writ large that the pieces of the puzzle are there, technically. If in that respect's any obstacle, it might comprise organizational: Microsoft's Office apps have been spun out into their own group, away from Cortana and Bing. Shum, however, said that intelligence agency is still part and parcel of Microsoft's offerings. "Perch assured that we are infusing AI technology into all Microsoft products," he said in Oct.

Microsoft representatives besides said that users should expect more from Microsoft in the future.

"We see value in conversations crossways a range of devices and experiences," Microsoft said in an October statement. "We're just at the beginning of what we believe is possible and sure as shooting see lots of chance to connect Cortana and conversations into a number of productivity scenarios. Today, Cortana integrates with Office 365 for peek-able information about approaching meetings, on with flight and package tracking, and Bing is also providing intelligent insights in real time in Office. We will continue to invest heavy here."

If Microsoft rightfully believes in productiveness, though, the future of speech recognition inside your PC probably isn't using Skype to book a hotel in Bangladesh. It's writing active the experience—but with your voice rather than your fingers.

Source: https://www.pcworld.com/article/410480/the-windows-weakness-no-one-mentions-speech-recognition.html

Posted by: salazarlaure1957.blogspot.com

Share this post