Wireless Phone Calls and Speech Production

There is a new viral video going around involving a “voiceless phone call”. Tom Simonite writes on NewScientist.com:

A neckband that translates thought into speech by picking up nerve signals has been used to demonstrate a "voiceless" phone call for the first time.

With careful training a person can send nerve signals to their vocal cords without making a sound. These signals are picked up by the neckband and relayed wirelessly to a computer that converts them into words spoken by a computerised voice.
[clip]
The system demonstrated at the TI conference can recognise only a limited set of about 150 words and phrases, says Callahan, who likens this to the early days of speech recognition software.

At the end of the year Ambient plans to release an improved version, without a vocabulary limit. Instead of recognising whole words or phrases, it should identify the individual phonemes that make up complete words.

I have no clue how this actually works (there’s an HMM in there somewhere, right?), but its implications for models of speech production ought to be significant. The folks over at Haskins Lab ought to be interested, I should think.

(HT Andrew Sullivan)

Here's the video. Cool stuff.

On Crowdsourcing and Linguistics

Rumbling around in my head for some time has been this question: can linguistics take advantage of powerful prediction markets to further our research goals?

It's not clear to me what predictions linguists could compete over, so this remains an open question. However, having just stumbled on to an Amazon.com service designed to harness the power of crowdsourcing called Mechanical Turk (HT Complex Systems Blog) I'm tempted to believe this somewhat related idea could be useful very quickly to complete large scale annotation projects (something I've posted about before), despite the potential for lousy annotations.

The point of crowdsourcing is to complete tasks that are difficult for computers, but easy for humans. For example, here are five tasks currently being listed:

1. Create an image that looks like another image.
2. Extract Meeting Date Information from Websites
3. Your task is to identify your 3 best items for the lists you're presented with.
4. Describe the sport and athlete's race and gender on Sports Illustrated covers
5. 2 pictures to look at and quickly rate subjectively

It should be easy enough to crowdsource annotation tasks (e.g., create a web site people can log in to from anywhere which contains the data with an easy-to-use interface for tagging). "Alas!", says you, "surely the poor quality of annotations would make this approach hopeless!"

Would it?

Recently, Breck Baldwin over at the LingPipe blog discussed the problems of inter-annotator agreement (gasp! there's inter-annotator DIS-agreement even between hip geniuses like Baldwin and Carpenter? Yes ... sigh ... yes there is). However (here's where the genius part comes in) he concluded that, if you're primarily in the business of recall (i.e, making sure the net you cast catches all the fish in the sea, even if you also pick up some hub caps along the way), then the reliability of annotators is not a critical concern. Let's let Breck explain:

The problem is in estimating what truth is given somewhat unreliable annotators. Assuming that Bob and I make independent errors and after adjudication (we both looked at where we differed and decided what the real errors were) we figured that each of us would miss 5% (1/20) of the abstract to gene mappings. If we took the union of our annotations, we end up with .025% missed mentions (1/400) by multiplying our recall errors (1/20*1/20)–this assumes independence of errors, a big assumption.

Now we have a much better upper limit that is in the 99% range, and more importantly, a perspective on how to accumulate a recall gold standard. Basically we should take annotations from all remotely qualified annotators and not worry about it. We know that is going to push down our precision (accuracy) but we are not in that business anyway.

Unless I've mis-understood Baldwin's post (I'm just a lousy linguist mind you, not a genius, hehe) then the major issue is adjudicating the error rate of a set of crowdsourced raters. Couldn't a bit of sampling do this nicely? If you restricted the annotators to, say, grad students in linguistics and related fields, the threshold of "remotely qualified" should be met, and there's plenty of grad students floating around the world.

This approach strikes me as related to the recent revelations that Wikipedia and Digg and other groups that try to take advantage of web democracy/crowd wisdom are actually functioning best when they have a small group of "moderators" or "chaperones" (read Chris Wilson's article on this topic here).

So, take a large group of raters scattered around the whole wide world, give them the task and technology to complete potentially huge amounts of annotations quickly, chaperone their results just a bit, and voilà, large scale annotation projects made easy.

You're welcome, hehe.

Jason Wins, hehe

As if it wasn’t obvious, I decided to reiterate Jason’s point from the previous post, regarding the ante-previous post by taking my post and running through Google’s English to Italian translation. A thing of beauty, haha. Enjoy:

Invece di commentare i miei commenters per quanto riguarda il mio post Blog di Amore, stile italiano, ho deciso di fare questo è un post --

In risposta a Jason's acerbic commento "Credo che la più grande macchina di traduzione è stato solo uno scherzo, la pubblicazione della traduzione automatica. :) ",

Con la presente risposta nel seguente modo:

Non essere talkin 'trash' bout mio prezioso Google traduzioni; senza di loro, non potrei mai leggere la mia e-mail amico spagnolo Ana invia. Il suo inglese è peggiore di quella di Google traduzioni, in modo I'll take Google (rimshot!).

E lei non crede che ci sia qualcosa di poetico nella prima riga. Ho potuto vedere alcuni 20th Century poeta americano Wallace Stevens iscritto come questo:

Abbiamo aspettato mesi e mesi
In attesa di Titlepage dolce,
Il sito dovrebbe offrire conversazioni
(E perché non parlare)
Ardente e appassionato editoriale
Le ultime notizie, un nuovo modello
Algonquin Round Table

On Google Translations

Instead of commenting to my commenters regarding my post Blog Love, Italian Style, I decided to make this it’s own post –

In response to Jason’s acerbic comment “I think the biggest machine translation joke was just posting the machine translation itself. :)”,

I hereby reply thusly:

Don't be talkin' trash 'bout my precious Google translations; without them, I could never read the emails my Spanish friend Ana sends. Her English is worse than the Google translations, so I'll take Google (rimshot!).

And don't you think there is something poetic in the first line. I could see some 20^th Century American poet like Wallace Stevens writing this:

We have waited months and months
In sweet Titlepage Pending,
The site should offer conversations
(and why not talk)
Passionate and fiery editorial
On the latest news, a new model
Algonquin Round Table

Blog Love, Italian Style

Sitemeter consistently shows referrals to my blog from the Italian language blog Taccuino di traduzione 2.0 which Google translates as Translation Notebook 2.0. Unfortunately, I lack Italian language skills, so I am unable to enjoy the blogs postings. But I thought I'd pass it along to any of you who may wish to indulge. The latest post has a great painting of the famed Algonquin Roundtable titled "A Vicious Circle" by Natalie Ascencios.

Meaning no offense to the superior original, but my lack of Italian drove me to Google translate the whole post. Reading this poor translation makes me want to run out, study Italian real quick, then read the rest of the blog:

We have waited months and months in sweet Titlepage pending, the site should offer conversations (and why not talk) passionate and fiery editorial on the latest news, a new model Algonquin Round Table, with videointerviste choirs, forums on different literary genres For readers who do not give up ever, a blog, reviews, reports, awards, cotillons and who knows what else.

All false promises. Although well prepared on the subject, the presenter (which surely read as a young Hamlet in jeans and black sweater, in some alternative theatre company) is uncomfortable in front of the camera (average training, anyone?), The writers guests look around terrified, set design probably is the work of a student to first weapons, the conversation is woody, boring and, above all, language, not to mention the editing of footage (used scissors?). A great sin. But this can only improve.

ciao

an ear for accents

This women has a gifted "ear" for accents. She starts in England, moves through Europe, on to Australia, then makes her way from west to east through North America. I'll note that her Texas and South Carolina are pure stereotype, but damn she nails California and Toronto.

(HT Andrew Sullivan)

"yeah right" again

Eureka! I posted about the prosody of the phrase "yeah right" some time ago here. In particular, I claimed there are 3 three interpretations of the phrase, but I don't have one of them in my dialect (Northern California), namely what I called "back-channel (sentiment agreement)" which is roughly equivalent to ‘mm-hmm’. However, I had no sound files. Now I've found a near perfect example of this mystery prosody in the trailer for Juno, about 36 seconds in (here).

You can also read my most excellent review of Juno here.