In article <[EMAIL PROTECTED]>, Fredrick P. Eisele Sr. <[EMAIL PROTECTED]> wrote:
> As internationalization proceeds I thought this might be a > good time to start a thread on the subject of the > "Universal Networking Language" (UNL). Are you familiar with UNL? Have you worked with it? Or did you just run into it on the net and consider it a nice idea? I could find little information under the mentioned URLs, but it appears that currently UNL is only a project some people would like to implement within the next 10 years. > [ description snipped ] This appears to be just another meta-language project. I don't see why this should turn out more successful than previous attempts which ran afoul of obstacles inherent in the structure of human languages: - Different languages grammaticalize different things. Some languages force every noun to be explicitly marked as singular or plural. Some don't. Some add a dual. Some languages force every verb to be marked for tense. Some don't. Some require additionally (or instead) marking for aspect. Etc. - Words don't have sharp meaning but rather cover an area of meaning. Sometimes these areas overlap, giving us synonyms. Between different languages, there are often no pairs of words that cover the same area of meaning. Instead, the area of a word from language A is covered by several different words in language B, and vice versa. What does that mean for the meta language? - It must grammaticalize all features grammaticalized in any of the target languages. This looks easy if you are only dealing with closely related languages like English or French, but if you add Russian or non-Indo-European languages like Arabic, Chinese, and Japanese, a large number of features need to be dealt with. - It must use semantic distinctions sharp enough to represent the meanings resulting from the intersections of all areas of meaning for the words in all target languages. >From this, two conclusions should be rather obvious: 1. An automatic translation from a meta language into a natural language is theoretically possible. However, the meta language will be very clumsy due to the need to represent an extraordinary amount of grammaticalized features and very fine semantic distinctions. 2. An automatic translation from a natural language into the meta language is impossible, since you need to *add* information that isn't there in the source. A human translator with his understanding of the text plus common knowledge can do this, but a machine can't. (Unless it's an AI with human equivalent understanding of a text. Which at this point in time is science fiction.) In short, the whole UNL project appears to be very idealistic, unrealistic, and at this time isn't remotely anything the Debian community should concern itself with. -- Christian "naddy" Weisgerber [EMAIL PROTECTED] See another pointless homepage at <URL:http://home.pages.de/~naddy/>. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
