In past columns, I’ve expressed my support for those people you’ve probably seen silently staring at their cell-phone screens, furiously typing away or intently scrolling. I recently joined their growing number. I think of it as a test drive; others might call it an occupational hazard. At any rate, I wanted to see what this wireless world is all about, and NTT DoCoMo’s i-mode seemed like the quickest ticket. (I know it’s not the only game in town, but the voluminous i-mode content makes DoCoMo king).
Even though I joined the nomadic i-tribe only a few days ago, I still find myself grinning at the thought of such a fast and easy Net connection, always on, always available. I’ve sampled online restaurant listings (with maps!), movie listings, club listings, and I’ve got much more packed into my pocket. So far, so good.
My brow usually furrows, though, whenever I actually input anything into my keitai, be it a Web address or an e-mail message. The thought of real-time text-chat gives me palpitations.
For so many mobile gadgets, higher portability means poorer input interface. The stylus and handwriting recognition solution (available on Palm PDAs, for example) is fine if you’ve got two hands and a little patience. The detachable mini-keyboard solution is nice if you have a lap or a table, and, in Japan, if you like kawaii pastel colors. Speech recognition holds promise, but it won’t be for everyone. (Just try whispering notes into your PDA/keitai during a board meeting.)
For now, it seems, the most accessible input method is thumb-tapping, which is akin to ballet in clogs. Flipping over to English, and then toggling between upper/lower cases is bad enough; jumping through the numerous kana conversion hoops of Japanese is migraine material. Sure, it’s not that different from ordinary Japanese word-processing, and apparently there are whiz-kids who can input with their eyes closed, but that doesn’t means we have to follow in their thumbprints.
Cell phones come with “time-saving” features such as pre-packaged phrases, ASCII emoticons and cute little symbols, and users have abbreviated language to speed up communication, but ultimately these workarounds will keep wireless communication in kiddyland.
One serious solution, which is unfortunately absent on Japanese cell phones, is T9 input technology. Developed by Seattle-based Tegic Communications (www.tegic.com) , T9 has practically become a global standard of cell-phone input.
To put it simply, T9 predicts what the user wants to say. If you want to type “hello” on a conventional cell phone, you have to hit the 4 button twice, the 3 button twice, the 5 button six times and the 6 button three times. That’s 13 taps for a five-letter word. Pity the fool who tries to dash off a quick “goodbye cruel world.”
With T9, however, to say hello you hit the 4 button once, the 3 button once, and by the time you hit button 5, the software has displayed a few possible matches, and continues to narrow down the choices with each tap.
To put it linguistically, which is what co-inventer Cliff Kushler did in an interview, the T9 algorithm “disambiguates” what a user inputs by calling up words from a database that correlate with grammatical context and their frequency of use.
Interestingly, T9 began as software to aid people with disabilities. A user wearing special eyeglasses would, via his/her gaze, navigate an eight-button key pad — the less effort, the better. Kushler and his colleagues optimized the letter layout to minimize the collision of frequent words, and later applied that same configuration to a cell-phone key pad. Statistically, the results had almost 99 percent accuracy.
“Being engineers,” said Kushler with mock pride, “we thought that was just great. We knew the world was ready for our nine-key keyboard that had random letters assigned to them. And we spent a lot of time knocking on doors, saying ‘see how visionary it is,’ but we found the world is not ready for such a key pad.”
The T9 software adapted for the conventional phone key pad has, however, been widely embraced. It is currently available on 50 handsets and in 14 languages, including Chinese (both simplified and traditional characters) and Korean. In December 1999, AOL acquired Tegic to enhance input for its instant-message software, which could be integrated into wireless PDAs and mobile phones. Also, we can expect T9 to appear on other devices, such as digital cameras and MP3 players.
Kushler said the Japanese version, which streamlines the input process of kanji and two sets of kana, not to mention romaji acronyms, has been Tegic’s biggest challenge — even after Finnish and Chinese. “When I first started the project, I wasn’t sure it was going to work,” said Kushler. Note this admission comes from a man who is fluent in Japanese, and who received his doctorate in engineering from the University of Tokyo.
The current Japanese version of T9, which was 21/2 years in development, boasts a database of 60,000 root words or word stems, which fall into 30 different grammatical categories. Add it all up and you’ve potentially got 300 million different simple phrases. Words can also be added to the database, which already contains 5,000 proper names.
The software’s ingenuity doesn’t stop there. Wanting to avoid a steep learning curve, the developers included an optional feature that observes the user’s behavior and offers suggestions. Another innovation was to fine-tune the kana-conversion process so the user doesn’t have to scroll through long kanji lists.
Tegic is still testing T9 in Japan. Why the global leaders of wireless communication haven’t snatched up licenses is a mystery to me. Kushler diplomatically explained that they’ve encountered “potholes and obstacles on both sides of any ocean you’d care to look at.”
In the meantime, keep your thumbs crossed.