Assume a good machine knowledge representation format exists. (An area of active research, I think.)
Assume a human user has the incentive to write directly into this format, rather than the typical "write in natural language, then use natural language processing to parse it into an internal KR format". (This kind of incentive already exists for programming languages, why not for other things?)
We will need significant tools: the UI problem seems very hard. The problem of choosing the coded representation of a concept is vaguely similar to input of kanji characters in Japanese. However, for Japanese, the user knows at a glance when the correct character has been found; in contrast, for inputting knowledge, it's harder to tell whether the UI's best guess is the correct one among many subtle shades of meaning.
No comments :
Post a Comment