Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!elroy.jpl.nasa.gov!ncar!ico!rcd From: rcd@ico.isc.com (Dick Dunn) Newsgroups: comp.society.futures Subject: Re: the interface for the rest of us? Summary: kanji Message-ID: <1991May3.205647.7392@ico.isc.com> Date: 3 May 91 20:56:47 GMT References: <9105021606.AA26962@lti2.lti.uucp> Organization: Interactive Systems Corporation, Boulder, CO Lines: 39 mh2f+@ANDREW.CMU.EDU (Mark Hahn) writes: > ...pens are practically a necessity for ideographic writing, like Japan. > currently, the state of the art for typing kanji is to enter > an English transcription, which is then converted into the Japanese > syllabic alphabet (kana) and finally into Kanji (the latter > is a many-many mapping.)... It does seem clumsy to our alphabet-oriented minds, but I don't think it's quite as bad as Mark makes it seem. With a Japanese keyboard, you don't need to start with romaji; the kana are already on the keyboard. That reduces the number of keystrokes by about a factor of two from what Mark explains. The typical action is to type two or three kana and hit a special "convert" key to get the kanji. The many-many problem Mark mentions is handled as follows: normally "convert" gives you the most-likely kanji for which you've just entered the pronunciation (kana). If that's not the right one, you hit "convert" again to get the next possibility. The con- venience of this method is critically dependent on the frequency with which the "right" kanji is presented first. Still, after all, it takes a few keystrokes to get a kanji, and a "word" as we conceive it is typically one or two kanji. The number of keystrokes to express a given thought should be comparable between kanji entered as above and English; it's certainly within a factor of two. On the other side, I would expect recognition algorithms for kanji to be pretty accurate, since there is a lot of convention for the order and direction the strokes are made...the problem of recognizing as it is written looks (from a naive view:-) to be considerably easier than an OCR problem. But I wonder what the actual ratio of writing:typing speed is for kanji... and I also wonder what future developments might help one or the other. Just for example, would contextual analysis help the probability of presenting the right kanji first in the keyboard method? (As far as I know, existing methods use strictly static ordering of kana->kanji mappings, although my info could be out of date. Anyone know otherwise?) -- Dick Dunn rcd@ico.isc.com -or- ico!rcd Boulder, CO (303)449-2870 ...If you plant ice, you're gonna harvest wind.