Method and means for grammatically processing a natural language sentence

A computer method for grammatically processing a natural language sentence made up of word units using processing means comprising the steps of:

• a. Lexicalizing each of said word units in accordance with verbal categories and application directed categories registered in a lexical memory, for storing the lexicalized word units in a word memory;
• b. Parsing said lexicalized word units read out from said word memory to provide at least one sentence representation for storing in a sentence representation memory for storing in a sentence representation memory, and including the steps of generating at least one constituent structure having functional indications from said word units, each of said sentence representations having associated therewith a probability factor and
• a) determining successively for each word unit supplied by the word memory and for each constituent supplied by the sentence representation memory and allocated with a constituted category the functional word category belonging to that word unit within the structure of said constituent by reference to grammar data list in a grammar data memory and concerning said verbal category of said word unit and the category of said constituent for updating or supplementing the respective sentence representations,
• b) describing for each constituent supplied by sentence representation memory and allocated with a category of a parent constituent, said parent constituent dominating said constituent, a closure process for said constituent based upon grammar data entered in the grammar data memory and concerning the category of said constituent and said parent constituent and (ii) allocating a functional label that can be provisional to said constituent to be closed by said closure process;
• and
• c) testing the interrelation between the words and constituents of said parent constituent as to functional categories of said words and said constituents during at least step b) (i) or b) (ii) against rules stored in said grammar data memory and based upon syntax and changing said probability factor if an error is detected and selecting each sentence representation having a probability factor above a preselected threshold value.