Walt Mossberg

A Phone That Takes Dictation: Testing Voice-to-Text Function

Cellphone text messaging, long popular abroad, is finally catching on in the U.S., especially among younger users. But cellphones are notoriously frustrating to use for entering text.

If you’re unfamiliar with using a numerical keypad to enter text (lucky you), spelling out a simple word like “dance,” for example, would require pressing 3, 2, 66, 222, 33. Typing out full sentences using this method is even more annoying, leading users to get creative with abbreviations and short-spellings. “R u goin 2 b l8?” is text-speak for “Are you going to be late?”

Phone makers have tried to solve this problem by squeezing little keyboards into the bodies of some phones. But these keyboards usually make phones bigger and bulkier than normal, and often show up only on costlier models, like the Treo or BlackBerry.

This week, my assistant Katie Boehret and I tested a new phone that attempts to solve the text-entry problem in a novel way that doesn’t involve typing, and can be used on a small, inexpensive phone with just a numerical keypad. This new phone lets you dictate your text messages by just speaking into the phone.

The Samsung p207, $79.99 with a two-year contract from Cingular Wireless, has built-in “speech-to-text” technology: It turns what you say into text on the screen. This technology, called VoiceMode, was created by a small Massachusetts company called VoiceSignal Technologies Inc. If it works properly, VoiceMode should make composing a text message as simple as dictating a voice-mail message.

Unfortunately, it doesn’t work very well. In our tests, the system made so many errors requiring tedious corrections that it might have been faster for us to peck out our messages the old-fashioned way — especially if we used the abbreviations and shorthand phrases so common among text-messaging fans.

The p207 is a handsome all-black flip phone with a built-in camera that, outwardly, betrays no sign of its special capability. To set up VoiceMode for the first time, you must teach your phone the acoustics of your speaking style. This requires reading 122 words out loud as they appear on the phone’s screen, and Katie did this in a few minutes.

Once the phone was trained for her voice, Katie set its text-messaging system to use VoiceMode by default; manual typing is still an option, you need only switch modes to type letter by letter again. The p207 also comes with VSuite, a more common technology from VoiceSignal that’s found in some 60 cellphones. It allows users to speak commands like “Call Walt,” instead of scrolling through menus to find phone numbers.

A special button on the phone activates VSuite and the voice commands. Katie simply pressed that button and said, “Send text, Walt Mossberg,” and she got to a blank screen for entering her text message, already addressed to me, since my phone number had been plucked from the phone’s address book.

Then, to dictate messages into the text screen using VoiceMode, you hold down the camera button on the side of the phone, which acts like a walkie-talkie switch. As you dictate, VoiceMode requires you to speak in a slow, unnatural manner, pausing after each word so the system can distinguish one word from the next.

The phone was trained to Katie’s voice, so she started off, speaking a rather simple sentence: “Hi (pause) period (pause) I (pause) am (pause) testing (pause) a (pause) new (pause) phone (pause) period.” But VoiceMode flunked even this easy test. It typed: “Hi. I am having a mail file.”

Once you release the walkie-talkie button, VoiceMode automatically corrects the words that it thinks were wrongly interpreted. While this helped during some of our tests, it usually didn’t correct sentences entirely. For Katie’s first test sentence, none of the words were auto-corrected.

samsung
The Samsung p207, $79.99 with a two-year contract from Cingular Wireless, has built-in technology from VoiceSignal Technologies Inc. that lets you dictate text messages. For more info: www.cingularwireless.com or www.voicesignal.com.

You can also edit words yourself, which Katie did by scrolling through her message, highlighting entire words and pressing the “0″ key, which pulled up a list of five words that she might have said, instead of what appeared on the screen. For example, Katie found “testing” on the list when she scrolled over “having” and selected it to make the change. This list also offers an option to spell out the word if you don’t see it on the list, by entering it the old-fashioned way one letter at a time. The whole process can be agonizingly slow if VoiceMode misses badly on the first try.

I had similar results when I tried my best teenager imitation. I said, “Hi. This is so cool. I’ll show it to you when we meet at the mall.” The phone wrote, “Hi. This is Saul cool. I’ll show it to you when we made at the walls.” To be fair, I hadn’t trained the phone for my voice, so it had a harder time recognizing what I was saying.

VoiceSignal says that the phone is constantly adapting, so as you go through the message and edit incorrectly detected words, it remembers those changes for the future. And indeed, the system did get better at recognizing some words the more we used it. But it was inconsistent, and messages still required significant editing.

The company says its dictionary holds “many tens of thousands of words,” including slang abbreviations and so-called emoticons — those little smiling or frowning faces created using punctuation. We tested that claim by saying “laughing out loud” into the phone; “LOL” — a common text message and e-mail abbreviation — appeared on screen. Speaking the words “smiley face” produced a “:-)” after three false results (“analysis,” “minimizing” and “violated”).

In addition to the frustrating inaccuracies of VoiceMode, there are other problems with speaking your text messages out loud. One of the benefits of sending a text message is that it can be done in sly silence, during boring meetings or classes. But dictation makes that impossible. It also makes your message public to anyone within earshot.

And unlike phone calls, text messaging doesn’t require a quiet environment; noisy bars and concerts are fine places to type text messages. But attempting to use voice-recognition technology in such environments might drive you crazy.

While VoiceMode has good intentions and did learn certain words as we used it more, we still found ourselves spending too much time editing our messages. This is one of those technologies that looks better on paper than it does in the real world.

With reporting by Katherine Boehret

Write to Walter S. Mossberg at mossberg@wsj.com


You’ve Come a Long Way, Silicon Valley

December 17, 2013 at 3:47 pm PT

Oh, the Places Your Phone Will Find

December 03, 2013 at 3:14 pm PT

Xbox One: Digital Home Base for the Living Room

November 26, 2013 at 3:25 pm PT

An iPad That’s Mini in Screen Size Only

November 19, 2013 at 3:06 pm PT

Latest Video

View all videos »

Search »

Walt Mossberg’s Product Guides

Desktop PC’s and Laptops

The Laptops to Buy


Digital Cameras

Digital Cameras Improve Zooms, HD Function