Voice test for autism

This is interesting. Speech researchers claim to have invented a voice test for autism. From the article:

“This technology could help paediatricians screen children for ASD to determine if a referral to a specialist for a full diagnosis is required and get those children into earlier and more effective treatments,” ‘The Daily Telegraph’ quoted lead scientist Prof Steven Warren of Kansas University.

Very nice. I love to see speech research put to practical use.

Dolphin gets an iPad

Merlin got an iPad...and the first thing he said was, "so long, and thanks for all the fish."

Pronoun haiku

A recent scientific study shows how the use of pronouns saves processing cycles in the brain.

And of course, if we didn't have pronouns no one could have written this excellent haiku:

I am he as you
are he as you are me and
we are all toge'er

Speech projects are special

The projects that I've worked on since 2002 have nearly all been implementations of speech recognition systems, or applied research on listener perceptions of speech reco systems. I've learned that speech reco projects have a lot of characteristics that distinguish them from typical software development projects. I've made note of those characteristics, and started to develop a lessons learned file to capture the kinds of things that must occur in order to ensure a successful project. I'll start posting those lessons over the next few months. Of course, if you have any favorite lesson learned from speech reco projects, please leave a comment.

(The title of the blog post is a take off on the "speech is special" argument that some linguists make based on phenomena like categorical perception that were first observed in speech experiments.)

Speech IVRs as hearing-impaired foreigners

I listen to recorded utterances from people talking to speech IVRs. One thing that people often do when when the IVR doesn't understand them is to change their speech pattern: they'll talk louder or more slowly. If the caller is soft spoken and there's noise in the background, then talking louder will help. Under most circumstances, though, shouting doesn't help, and talking more slowly almost never helps.

If you've ever traveled abroad (or live outside the US) then you may have witnessed US tourists trying to make themselves understood to non-English speaking locals. Often, the tourists will shout and talk more slowly, adding helpful gestures in an amusing pantomime, in order to make themselves understood. In fact, adding gestures can help, and speaking more slowly can help if the local speaks some English, and if it isn't done to the point of changing one's pronunciation.

Having witnessed these two scenarios many times here's my theory of the day: callers to speech IVRs believe the systems to be hearing impaired foreigners who are best communicated with by shouting and speaking slowly to. If you were to watch these people on the phone I'm confident that you'd see the pantomime routine as well.

The passing of Alex, and the "Alex Rule"

Alex, the African Grey Parrot who was said to be able to use language, died last week of unknown causes. The announcement of his passing was delayed in order to give researchers time to "get over the shock and talk about it."

I became aware of Alex in the mid 1990's when I was doing research on spoken word perception. A visitor to the lab where I worked described meeting Alex, who was already famous among language researchers. She was required to wear sterile surgical scrubs before entering the room where Alex was kept. We lab workers were very impressed with our visitor and her story of having met near royalty in the language world.

One of the issues regarding Alex's use of vocalizations was whether the bird was using language in a human sense. That is, did Alex use words and word order (syntax) in order to express ideas? Selective portions of written transcripts of conversations between scientists and Alex appear to show language use at about a three-year-old level. A portion of a conversation is found in this article in Scientific American.

The really compelling examples of Alex's language skill are the recorded conversations. It's hard for people to see and hear these snippets of interactions and not think that the bird is using language in a human way. Maybe we're wired this way, to accept human vocalizations as expressing human-level emotion and thought.

In a way, our speech IVRs are a little like Alex. They're programmed to produce human vocalizations and recognize and act on snippets of input ("utterances" is the preferred term, instead of words, phrases, or sentences). From this very meager verbal repertoire the humans who call and talk to the systems tend to ascribe capabilities to the system that don't exist. This shows up in callers' generating long, verbose utterances, and courtesy language, and descriptions of such systems as "smart" or "stupid," depending on the systems' performance. Unlike Alex, however, most speech IVRs are neither cute nor engaging, and people who talk to them are usually just trying to get a request fulfilled and get on with their day. Designers who "personify" their VUIs-that is, name them and embue them with personality-are usually making a mistake.

To designers of VUIs I submit the Alex Rule: Don't personify your VUI unless you can make it as charming as Alex. You can't create a VUI as charming as Alex, so don't try.