| 摘要 |
A method and apparatus for processing natural language and deducing meaning from a natural language input characterized by the steps of (a) receiving an ordered string of word objects having a natural language meaning, (b) selecting a word window length, and (c) successively moving the word window along the ordered string and analyzing the meaning of a substring of word objects that fall within the word window. The substring is removed from the ordered string if the substring has a recognized meaning, until all substrings of the ordered string that fit within the window have been analyzed. In a step (d), the word window length is reduced and step (c) is repeated until only an unrecognized residual of the ordered string remains. The meaning of the substring is analyzed by mapping the substring against a database using one or more mapping routines. The mapping routines are preferably arranged in a hierarchy, wherein a successive mapping routine is used to analyze the substring when a previous mapping routine in the hierarchy cannot map the substring. A computer-implemented task is determined from the recognized substrings and performed by the computer system. The apparatus of the present invention implements the method on a pen-based computer system, and the ordered string is preferably received from strokes entered by a stylus on a display screen of the pen-based computer or from a microphone receiving speech input.
|
