Slashdot Mirror
Microsoft Speech Recognition Now As Accurate As Professional Transcribers (techcrunch.com)
An anonymous reader quotes TechCrunch:
Microsoft announced today that its conversational speech recognition system has reached a 5.1% error rate, its lowest so far. This surpasses the 5.9% error rate reached last year by a group of researchers from Microsoft Artificial Intelligence and Research and puts its accuracy on par with professional human transcribers who have advantages like the ability to listen to text several times. Both studies transcribed recordings from the Switchboard corpus, a collection of about 2,400 telephone conversations that have been used by researchers to test speech recognition systems since the early 1990s. The new study was performed by a group of researchers at Microsoft AI and Research with the goal of achieving the same level of accuracy as a group of human transcribers who were able to listen to what they were transcribing several times, access its conversational context and work with other transcribers.
On a daily basis in my work environment Microsoft technology is used to a) record voicemail and b) generate text from the speech. Never, ever, have I received any converted voicemail that wasn't completely unintelligible gibberish. Seriously. This is utter nonsense.
When a human transcriptionist makes a mistake you can usually work out what they meant. When Speech-to-text (STT) makes a mistake it is often gibberish. So objectively it is "better" at transcribing, but subjectively much worse.
Or did the software get a filtered recording (or signal processes it itself) and the humans an unfiltered noisy static recording?
How about we try with high quality modern voice recordings instead of 20+yr old telephone line recording?
holyfield is these all of this was made worse by the fact that i had these birds skilled estimate uh... supplying itself what's your special prom to prevent fraud reform
thoughtfulness julia roberts police comments entry drug connections predicting that nighttime beating
Harald
Actually speech goes to the cloud and is translated by lots of MSCEs they'd otherwise lay off!
Some months ago, I did some tests with speech recognition software and my conclusion was that it is still too unreliable. My intention was to develop an application allowing me to write moderately complex code by voice (creating files and folders, including proper indentation, recognising functions, variables and other basic elements, etc. Basically, allowing me to write/edit the main parts of a random algorithm in certain language without touching the keyboard). I did test Microsoft in-built functionality (+ used one of Microsoft's .NET programming languages) and it wasn't even close to what "5.9% error rate" seems to indicate (almost perfect?).
In defence of the software, I have to say that my English accent isn't precisely excellent (some people say that it is "too thick" and other people just say "what?". LOL) and honestly I make a very little effort to pronounce properly. But this is also the problem with speech recognition: it is mostly focused on a specific language/accent/intonation. I was doing my tests in an English Windows version and this was the language for the default speech recognition (and adding a different one wasn't precisely straightforward).
I do perfectly understand the complexity associated with developing a reliable enough piece of software delivering what I was expecting; but this is precisely the reason why I looked for existing solutions rather than developing everything myself (what I do pretty often). In any case, my impression is that you can still not expect good enough reliability of (Microsoft's) speech recognition software, much less when mixing languages/accents up (particularly problematic situation: including Spanish words when talking in English). I might give a new shot at all this next year though.
Custom Solvers 2.0 = Alvaro Carballo Garcia = varocarbas.
of your lovely data is slurped by MS during the translation?
All of it? Now thats a surprise.... Not...
"As Accurate As Professional Transcribers..."
They left out "from Uzbekistan transcribing Navajo - underwater".
Never trust anything Clippy say.
Microsoft piece recognition is a cod am pizza ship.
Microsoft socks!
General Protection fault...
Dear aunt, let's set so double the killer delete select all.
Does this mean it is good, or does it mean they overtrained it on the samples and those people in those recordings are the lucky few it will understand now?
They should do tests using modern hardware. For example the speech recognition on iOS seems to be pretty good. If they can get this technology into windows 10 that would be awesome. Oh I dictated this using iOS.
The NSA would love this. Keyword scanning of 95% of what's spoken in phone conversations (given enough processing power to transcribe them all).
Better known as 318230.
Just make sure you run it on an air gapped computer if you want your conversation to remain private.
Seven puppies were harmed during the making of this post.
That means :
- they published against a corpus from the 90`s... Not real world results (ya know, ground truth and sciency and all)
- it's, ya know, research... they side with publish or perish mindset...
- if this was so amazing, the *product* team would be claiming it. And, given the viable *product* they have is Windows, the chances for tech transfer to Microsoft products are near zero.
So treat this like some college PhD thesis paper, not a pre-product announcement. Must such for Microsoft research, knowing that their results won't make it into products.
I worked as a professional transcriber in the legal profession, actually employed by the government. 95% accuracy would be 1 mistake in 20 words, an error almost every 2 lines. For the standard we had to type to, an error every 2 pages would be unacceptable. These transcripts are admissible evidence in court as an exception to hearsay rules and people's lives hang on the accuracy of them. The transcripts themselves are also literally the law of the land (I live in a common law jurisdiction, so my transcript is literally legally binding law and a printout of my transcript is admissible for that purpose as well). Imagine a 5% error rate in that.
Also, judges always speak "The Queen's English". How is this algorithm going to translate what they really say into proper language suitable for a judicial order? I'd also love to see how it deals with technical Jargon; for example citations that are spoken all sorts of haphazard ways yet must be typed in a specific format.
And this doesn't even factor in the thick accents many people use that are almost unrecognizable by the best humans, how is a computer going to deal with that?
Excellent job, as always.
At work we have an cloud-based Outlook that transcribes voicemail to text. It's so comically inaccurate that we sometimes forward the results to the sender and we both get a good laugh.
Competition Good, Monopoly Bad.
If it can recognize "It's difficult to wreck a nice beach", I'll be thoroughly 'whelmed'.
The lameness filter is lame.
Holyfield is that all this was worse than the fact that this bird expert was appreciating uh ... providing the same as his own promotion to prevent fraud reform Reflection Julia Roberts Comments Police links Drug entry Expect that night beatings
This could save doctors a ton of money if this is true. I suppose transcribers are for teenagers and not for raising a family and they should have made better life choices than demand $15/hr?
Even if there are higher error rates you can have 1 real human monitor 50 computer automatized transcriptions where the computer is fairly sure but not 100% certain. Hospitals are always looking for ways to save pennies
http://saveie6.com/
IgPay AtinLay?
Peace is easy to achieve, just surrender. Liberty is much harder get/keep.
puts its accuracy on par with professional human transcribers who have advantages like the ability to listen to text several times
Is this a joke or does the author have literally no idea how computer memory works?
In a sound proof studio built for sound recording spoken by someone with speech training?
Or in an environment with 30 people talking in the background, an air condition running, doors and drawers slamming, people laughing, feet
and chairs shuffling across the floor, some photocopiers that got their last service before Bush left office whining for hours and a person speaking into the phone while at the same time talking to coworkers and you're expected to know which words belong to you and which ones are directed at someone else?
Aka "open plan office".
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
Way to miss the point. Automation doesn't mean that the task has to be fully automated. Even just partially automating a task can have huge savings.
Let's look at the situation you've described. Instead of having let's say 10 highly-trained transcriber doing the transcriptions manually, they'd have 2 less-trained and/or less-skilled (read: cheaper) workers verifying the transcriptions done by computers and making corrections where necessary.
So the bulk of the work is offloaded onto the computer, and a smaller number of cheaper humans are used to detect and fix up the 5% to 10% of mistakes made by the computer.
So instead of paying 10 humans a high wage, they're paying 2 humans a lower wage. Maybe the 10 humans cost $1,000,000 a year. Now they're only paying $60,000 a year in wages, plus the negligible cost of running the computer software. It could be an annual savings of 90% or more, with just partial automation.
Microsoft Speech Recognition Now As Accurate As Professional Transcribers who are deaf and whose native language is Esperanto.
It still showed up at the South Park "Save Films from their Directors" club for the wrong reason when it heard, "Free Hat".
(For those that aren't South Park followers...)
Cartman writes "Free Hat" on the advertising poster in the belief that freebies are necessary to attract people. However, the crowd mistakenly thinks the rally is to free Hat McCullough, a convicted baby killer they believe was innocent.
Now thinking that "Free Hat" would be a great name of one of those Windows App Store pirate streaming apps ...
It must have been something you assimilated. . . .
I was pleasantly surprised by the voice-message to email service my last employer had with Google.
They sent you the voice message in an attachment with the translation in the email. If the translation didn't make sense, you could play the audio yourself.
Only annoying thing was we still had to delete the VM off the phone manually afterward.
Will it transcribe, "Diffused the situation," or "Defused the situation"? Every single TV closed-caption I've ever seen, and I've taken special note since I first became aware of this, has gone with the former. And those presumably have been humans making that error.
If you believe Microsoft without independent verification from an otherwise uninterested third-party who has no investment in the outcome, then you're a fool.
One in 20 words is wrong?
How can a human transcriptionist be that bad?
Help! I'm a slashdot refugee.
Native speakers in a southern court room.
I bet you also want me to believe that Jesuits survived Hiroshima because of their rosaries.
Good results on a small corpus are no guarantee of future performance. MS should provide a website where you can talk to their system and see what it outputs. Then, we'll be separating the men from the covfefes.
Humans transcribers "have the advantage to be able to listen to the recording several times"? What utterly demented nonsense is that? Of course, the software, having the recording, can "listen" to it as often as it wants. There is absolutely no "advantage" here for the human transcribers.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
As if the audio sails by the program and isn't stored in memory and parsed as many times as needed.
I fuse micro sot noise recognition ball the time it words fall Leslie.
The Russians have won. They have made the world a cesspool of distrust, greed, fear and hate.
The Miracle of Hiroshima -- Jesuits survived the atomic bomb thanks to the rosary
It is? And who decided *that*?
We've got it on our hybrid phones. At least half the time, the voice transcription "preview" resembles, randomly, Vogon poetry, or perhaps only "computer poetry" from 40 years ago. It rarely gets a name or title correct, and the message they're trying to leave, *maybe* 50% is close enough to guess what they meant, without listening to the mp3.