PRINT December 1985


Duchampian chance and poetry in Japanese-language word processors.

THE MOST FASCINATING part of the system of thought known as information theory, devised in the ’40s by the American mathematician Claude Elwood Shannon, is the idea that information is transmitted more accurately if some degree of noise and redundancy are incorporated into the message. The extra noise may not be essential to the message’s meaning, but it is essential in conveying the meaning. An example of the theory in action is the handling of the deep-space craft Voyager II by NASA engineers, who arranged for “100 percent redundancy” in its transmission of images of Saturn; in other words, for each bit of information the ship sent, it deliberately added another. The information was like a fish colony swimming in a sea of noise.

A ratio between information and redundancy is in fact implicit in the human consciousness, not to mention in its creative activities. Most fine art works with a high redundancy ratio. Most machines, whose reason for being is efficiency, are at the opposite end of the scale, while language lies at the midpoint between the two. Language usually digests a “raw” message only with the aid of a fair degree of redundancy, and the extra noise not only helps convey the message but also introduces the ambiguity that is the driving force in sustaining language’s freshness. It would be difficult to design a machine that could maintain this aspect of language as living information. It is hard to imagine an English-language or French-language word processor whose mechanical system was ambiguous in itself, making it unnecessary to input redundancy from the outside. In Japan, however, word processors have somehow made room within themselves for redundancy to nest.

To write Japanese we resort to two separate systems: the kanji inventory of over 10,000 ideogrammatic characters, taken from China, and two sets of domestically developed alphabets, each with over 50 signs, called kana. For example, “flower” in kanji can be represented as “” but it can also be “ハイ” or “はな” in kana. Kana characters are phonetic; they represent the sounds of words rather than their meanings, and as a result kanji symbols, which, being ideograms, do carry meanings and ideas, are often preferred. For everyday use one only has to know by rote about 500 kanji characters, but even for this reduced inventory it has proved impossible to develop a user-friendly, 500-key, typewriter or word processor keyboard. Even a technology-minded pianist would be unable to tackle such a monster. So a word-processing system was developed that uses a kana keyboard of about 50 keys. The kana characters must be combined in specific ways to result in kanji symbols.

The first exploration of the system was in the ’60s. In its initial stages one used kana characters to punch a Japanese text into a long thin paper tape. The tape was read by a computer and then converted, through batch processing, into a text that mixed kanji and kana symbols together. Here the computer functioned as a dictionary to connect a set of kana to its proper kanji. In the ’80s, from the development of systems such as the Toshiba company’s JW10, a new species of word processor emerged, which yields kanji output from kana input immediately through the use of a dictionary-function floppy disc carrying 5,000 kanji symbols. But these systems had an unexpected and interesting byproduct. If you type “ハイ” one of the phonetic characters for “flower,” the machine may print either “” or “” in kanji characters. Phonetically these characters sound the same, yet the former means “flower” and the latter “nose.” The phenomenon arises from the abundance of homonyms in Japanese. The machine has inherited the ambiguity of the language, its master.

Working with one of these word processors the user must continually select the required kanji symbol from a group of possible candidates that the machine posts on the screen, and this selection process has become an integral part of literary production for a Japanese writer of the electronic age. The writer is constantly made aware of the relationship between phonetics and semantics in the language, made aware that a word is an ephemeral entity At the same time, the machine is evolving too. New programs allow the word processor to learn for itself about the selective nature of the conversion procedure between the two systems of characters. The WD series put out by the Sharp company for example, is equipped with what is called “prioritization.” An English analogue of the function might run as follows: suppose you input the word “flower,” you could arrange that the next time you type just the letter f, “flower” would appear on the screen as your top-priority choice. This kind of programming gives each machine a kind of equivocal postnatal individuality.

The ambiguity of Japanese language word processors stems from the ambiguous structure of information expression in the language itself. At the same time, it implies the possibility of incorporating the essence of ambiguity in a mechanical system. If exploration of this direction continues as a legitimate effort, we may someday see an almost science fiction—like situation in which a language machine would have the function of hesitation that has long been monopolized by the arts. In the foreseeable future, teenagers may build and destroy structures like Marcel Duchamp’s Large Glass, 1915–23, with its in-built operations of chance, on word processors as daily routines.

Seigow Matsuoka is the editor-in-chief of object magazine YU and freelance editorial director who lives in Tokyo. He contributes a column on philosophies of craft to Artforum.

Translated from the Japanese by Kazue Kobata