초록 ▼

Apparatus and method are provided for segmenting, parsing, interpreting and formatting the content of instructions such as air traffic control instructions. Output from a speech recognizer is so processed to produce such instructions in a structured format such as for input to other software. There

Apparatus and method are provided for segmenting, parsing, interpreting and formatting the content of instructions such as air traffic control instructions. Output from a speech recognizer is so processed to produce such instructions in a structured format such as for input to other software. There are two main components: an instruction segmenter and a robust parser. In the instruction segmenter, the recognized text produced by the speech recognizer is segmented into independent instructions and each instruction is processed. The instruction segmenter receives a recognized air traffic control or other instruction, and segments it into individual commands. Utterances or other language are thereby broken up into their component instructions by detecting probable instruction boundaries. If normal processing fails, then robust backup processing is invoked, as a fallback after a fixed time per word has elapsed or a processing failure has occurred. The robust parser allows the system to extract information from utterances that are not necessarily well formed or may have extraneous comments in them.

대표청구항 ▼

A computer-implemented method for robust language processing, comprising the steps of: accepting language in text form; segmenting the text language into one or more individual instructions, said segmenting being performed in accordance with predefined sets of rules and specific words that indicate

A computer-implemented method for robust language processing, comprising the steps of: accepting language in text form; segmenting the text language into one or more individual instructions, said segmenting being performed in accordance with predefined sets of rules and specific words that indicate at least the beginning or end of an instruction; parsing the one or more individual instructions; and skipping one or more words in an instruction if a parse cannot be accomplished by said parsing step in a predetermined time or without a parsing failure; wherein said parsing step and said skipping step together comprise the steps of: first determining whether parsing is presently at any of at least one predetermined keyword; if the portion of the instruction being parsed is not at any of the at least one predetermined keyword, then moving parsing ahead one word and repeating said first determining step; if the portion of the instruction being parsed is at any of the at least one predetermined keyword, then parsing the instruction and second determining whether the end of the input has been reached; and if the end of the input has not been reached, then repeating said first determining step.