Language Log

From Diane Moderski:

An acquaintance of mine has a new iPhone, which he carries in a pocket that is (relevantly) below waist level. He has discovered something that dramatically illustrates the difference between (i) responding to speech and (ii) responding to speech as humans do, on the basis of knowing that it is speech.

The "silly AI doing something stupidly funny" trope is a powerful one, partly because people like to see the mighty cast down, and partly because the "silly stupid AI" stereotype is often valid.

But as with stereotypes of human groups, the most viral examples are often fakes. Take the "Voice Recognition Elevator"  skit from a few years ago, which showed an ASR system that was baffled by a Scottish accent, epitomizing the plight of Scots trapped in a dehumanized world that doesn't understand them. But back in the real world, I found that when I played the YouTube version of the skit to the 2010 version of Google Voice on my cell phone, it correctly transcribed the whole thing.

And I suspect that the recent viral "tuba-to-text conversion" meme is another artful fraud.

My dad accidentally texted me with voice recognition…while playing the tuba

(h/t Chris Waigl)

[Update: Mark Liberman suggests this might be some artful fakery. See: "Another fake AI failure?"]

In all tech considered (8/24/16), Arrti Shahani has an article titled "Voice Recognition Software Finally Beats Humans At Typing, Study Finds".

Turns out voice recognition software has improved to the point where it is significantly faster and more accurate at producing text on a mobile device than we are at typing on its keyboard. That's according to a new study by Stanford University, the University of Washington and Baidu, the Chinese Internet giant. The study ran tests in English and Mandarin Chinese.

Baidu chief scientist Andrew Ng says this should not feel like defeat. "Humanity was never designed to communicate by using our fingers to poke at a tiny little keyboard on a mobile phone. Speech has always been a much more natural way for humans to communicate with each other," he says.

Until two days ago, I had never heard of this word — even though I knew about Punch and Judy shows.

From Wikipedia:

A swazzle is a device made of two strips of metal bound around a cotton tape reed. The device is used to produce the distinctive harsh, rasping voice of Punch and is held in the mouth by the Professor (performer) in a Punch and Judy show.

Swazzle can also be pronounced or spelled Schwazzle or swatchel.

I like the fact that the performer is called "Professor"!

In late 2012, while visiting my son Tom in Dallas, I noticed that he was doing something very odd with his cell phone.  Most people enter text into their cell phone by pressing their thumbs (or their fingertip) on the letters of a small keyboard, whether virtual or actual.  But Tom was doing something altogether different:  he was sliding his finger over the glass surface of his phone and somehow, by so doing, he was able to enter text.  I was dumbfounded!  What amazed me most of all was how casual he was about it.  He'd be talking to me about something, then glance down at his cell phone, move his fingertip around on the glass, and — presto digito! — he'd have typed a message to someone and sent it off.

Baidu ("the Chinese Google") is a popular search engine in China.  The web services company (registered in the Cayman Islands) and its name are discussed in "Soon to be lost in translation," which I posted a little over a year ago.

Now Baidu has launched a new machine translation service.  A friend of mine in China impishly suggested that I give Baidu Fanyi a whirl by typing in 我恨中国.  Language Log readers are invited to try it themselves and see what they get.

A couple of days ago, Gary Marcus told me about the Beatles Complete on Ukulele project, and introduced me to its creator, David Barratt.

Gary got involved because he's working on a book about "learning to become musical at the age of 40", and so he's joining a roster of performers that includes the Fort Greene Childrens Choir (Age 7 and Under Section), Samantha Fox, and many others (82 so far), recording voice-and-ukulele versions of all 185 songs in the Beatles catalog. Gary is of course singing With a Little Help from My Friends (because, he explains, "otherwise I couldn't carry a tune in a bucket"), and his contribution is scheduled to be released on July 19, 2011.

So how does Language Log come into this? Well, David wants to recruit Barack Obama to sing Let it Be, and Gary thought that I could help. In turn, I believe that YOU can help.

The most recent IEEE Signal Processing Society Newsletter has an interesting article by David Suendermann, "Speech scientists are dead. Interaction designers are dead. Who is next?".

His argument is that "Commercial spoken dialog systems can process millions of calls per week", and therefore "one can implement a variety of changes at different points in the application and randomly choose one competitor every time the point is hit in the course of a call", using techniques like reinforcement learning to adaptively optimize the design. As a result, "the contender approach can change the life of interaction designers and speech scientists in that best practices and experience-based decisions can be replaced by straight-forward implementation of every alternative one can think of".

I wouldn't have predicted that this would work as well as it does:

Under the title "English without Chinese at exams 'traitorious,'" China Daily ("China's Global [English] Newspaper") presents an article by Wu Yiyao describing the uproar over the decision by four Shanghai universities to include an English test as part of their independent admission examinations, but not to include a corresponding examination for Chinese language. The controversy swiftly spilled over into other media reports, with strong opinions on both sides.

Top story of the morning in the UK for the serious language scientist must surely be the report in The Sun concerning a children's toy mouse that is supposed to sing "Jingle bells, jingle bells" but instead sings "Pedophile, pedophile". Said one appalled mother who squeezed the mouse, "Luckily my children are too young to understand." The distributors, a company called Humatt, of Ferndown in Dorset, claims that the man in China who recorded the voice for the toy "could not pronounce certain sounds." And the singing that he recorded "was then speeded up to make it higher-pitched — distorting the result further." (A good MP3 of the result can be found here.) They have recalled the toy.

Shocked listeners to BBC Radio 4 this morning heard the presenters read this story out while collapsing with laughter. Language Log is not amused. If there was ever a more serious confluence of issues in speech technology, the Chinese language, freedom of speech, taboo language, and the protection of children, I don't know when.