Dear Prof. Genenz, > From my database with roughly 50.000 lexical > entries (compounds) I get a > number of 1431 compounds with > at least two readings and 71 with at least three > readings. Taking only into account compounds with multiple readings.
But imagine this: If a program had merely access to a database containing the readings of single characters, it still couldn't figure out the reading of a compound (reliably). How could it "know" that 人間 is "ningen" and not, for instance, *jinkan? This means that it would be vital to have access to a database containing entries for kanji-compound->reading(s). Without semantic analysis of the sentences concerned, it is not possible to determine the correct contextual reading of every Japanese compound. It is only possible to check whether the reading given in the second string is _one_ of the correct readings. Greetings from Edo, Berthold Frommann __________________________________________________ Do You Yahoo!? Yahoo! BB is Broadband by Yahoo! http://bb.yahoo.co.jp/
