Lexicus Debuts Chinese Recognizer

Chinese Handwriting System Offers Impressive Performance

Despite certain studies that seem to indicate only marginal productivity increases related to computer automation, it’s clear that computers have changed the nature of the economies that have been able to exploit their benefit. Computers enable airlines to carry more passengers, manufacturers to employ advanced robot-controlled welders, and experts to publish industry-specific opinion and analysis on paper and on the information superhighway. Computers have been a great boon for societies whose communication matches the form used by present-day devices. By and large, this has meant either western societies, or others that employ a reasonably-sized alphabet. But as we know, not all societies are based on this form of communication.

This point has been driven home especially in the past decade as the Asian-Pacific region has raced ahead of the established world markets, outstripping nearly everyone with break-neck growth and opportunities. No longer confined to the four so-called “Tigers” consisting of Taiwan, Hong Kong, Singapore, and South Korea, economies from Thailand to the newly liberalized Vietnam are redefining the world economic balance. Add to this perhaps the most important emerging economy of all—China. Following liberalization that began in 1978, the People’s Republic of China (PRC) is now ranked as the second largest economy in the world when measured by the more accurate price point parity index, or even the Economist’s more whimsical Big Mac index.

With foreign investment inflows measured in the tens of billions of dollars, the mainland Chinese are eager to automate. Unfortunately, this has meant adopting computing systems that are ill-suited for their style of communication, which uses thousands of individual characters. With a single country accounting for a fifth of the world’s population, if there ever was a place for pen computing, this is it. Recognizing this vast opportunity, Motorola’s Lexicus Division has turned their considerable talent towards this problem, culminating in the Fall COMDEX introduction of the company’s Chinese Handwriting Recognizer.

“Accurate handwriting recognition should propel the use of computers in the very sizeable Chinese-speaking world.”

While not the first recognizer designed explicitly for Chinese recognition, Lexicus’ product has two important distinctions: it’s state-of-the-art technology, and it’s marketed by a leading company already well-established in the Chinese market. The Chinese recognizer is designed to be used without training, accepting input in either the Traditional or Simplified character sets. Traditional characters, of which the recognizer accepts 13,000, are primarily used in Taiwan and Hong Kong. The recognizer is also capable of understanding 6,000 Simplified characters, popular in the PRC and Singapore. In fact, this is one of the strengths of the Lexicus recognizer. Writer can select to have their characters recognized in either of the two writing styles, essentially “translating” from Traditional to Simplified, or vice versa.

This option could prove valuable as intra-regional trade expands. As Derek Ling, Lexicus Business Development Asia Pacific, noted: “There is so much communication among these countries now, and with 1997 coming with respect to Hong Kong and China, more and more people will need to communicate with the other side. What our program allows is for people to write Traditional characters and have it output Simplified, and the reverse; have people write Simplified and have it come out Traditional.”

Cursive Assumes New Meaning

Chinese characters consist of several closely drawn strokes, ranging in number from one to seventeen. The concept of cursive exists in written Chinese, but not the style to which western writers are accustomed. Instead, Chinese “cursive” has the strokes of a character connected, making the task of recognition just that much more difficult. But according to Ling, “basically this is what allows Chinese people to write in a natural way—instead of using keyboard input with any of a number of methods, one of which is romanization.”

In the version demonstrated at COMDEX, Lexicus used a piece of middleware called Chinese Pro. However the shipping version of the Chinese Recognizer, due in the first quarter of 1996, is intended to work with Windows 95 Chinese version which itself should be available in December. This enables a writer to input the recognized characters into any Windows application.

Based on its own tests, Lexicus is claiming an accuracy rate in excess of 95% based on real-world scenarios. The recognizer offered impressive performance on a 486-class device, easily keeping up with the fast-writing Ling, due in part to having much of the recognition occur in the background while continuing to accept new written input. The recognizer includes several attractive features including the ability to quickly select from among several of the program’s best alternative guesses when an unintended character is returned. However, the current version of the Chinese recognizer does not allow training to expand the dictionary or improve recognition. Ling stated that this is a likely enhancement in a future release.

New Market, New Technology

Lexicus is well known for their neural network-based Longhand recognizer, which incidentally also received a COMDEX-announced enhancement (see Short Cuts in this issue for more information). But according to Ling, the Chinese language was better suited to another approach: “It’s a new technology. The problems and issues are different. We’re not dealing with recognizing an alphabet of 26 characters, but a lot of characters.”

Motorola is also likely to find that the new market requires an approach unlike others in the world. When asked about the focus on an emerging market as opposed to a more established regional power such as Japan, Ling cited two important factors: “Motorola has a great presence in China, that’s one of the reasons. Also, there are a lot of people already working in the Japanese market.” With its proven track record in wireless technology markets in China, Motorola is almost uniquely positioned to deliver the fastest growing large economy in the world its ticket into the world of widespread computer automation.

Motorola, Lexicus Division
490 California Ave., Suite 300
Palo Alto, CA 94306
Wako Takayama
(415) 462-6800
(415) 323-4772
[email protected]