Molecular natural language processing system

• US 5,297,040 A
• Filed: 10/23/1991
• Issued: 03/22/1994
• Est. Priority Date: 10/23/1991
• Status: Expired due to Fees

• Current Assignee: Her Majesty The Queen In Right of Canada As Represented By The Minister of National Defence
• Original Assignee: Franklin T. Hu
• Inventors: Hu, Franklin T.
• Primary Examiner(s): McElheny, Jr., Donald E.
• Application Number:
  US07/781,761
• Time in Patent Office: 881 Days
• Field of Search: 364/419.1, 364/419.02, 364/419.04, 364/419.08, 364/419
• US Class Current: 704/9
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique

Abstract

A natural language processing system for processing digitally encoded text by utilizing a "molecular" method. This method combines words into sentences like atoms combine to form molecules. Each word in the language is treated like an atom which has particular attributes and binding characteristics. By verifying if these binding characteristics have been satisfied, it can be determined if a set of words can follow each other to form sentences of the language.

Claims

1. Apparatus for the analysis of digitally encoded natural language, such apparatus comprising

means for receiving encoded natural language text for processing,

dictionary means for storing words of the natural language in which each word in the dictionary can have a rule which specifies the types of words or phrases that can follow each word in the dictionary and their syntax, processing means for analyzing a sentence by verifying that the next word or phrase in a sentence is allowed by the rules contained in the previous words.

2. Apparatus according to claim 1, wherein said dictionary contains data for each word comprised of:

the word token,

the semantic and syntactic attributes,

a rule which expresses the possible syntactic constructions which can follow the word token,

a pointer to an alternate definition.

3. Apparatus according to claim 1, wherein said dictionary contains data for phrase structure rules comprised of:

the phrase structure rule token,

the attributes of words which can start the phrase structure rule,

the phrase structure rule expansion,

the inheritor of any attributes.

4. Apparatus according to claim 1, wherein said rules which indicate the types of words and phrases which can follow a word are expressed using a syntax which uses:

bracket symbols "<

" &

">

" or other symbols or methods to delineate a token of the rule,

an exclamation point "!" or other symbol to indicate that the token is a required part of the syntax,

a slash "/" or other symbol to indicate that either of the tokens which are adjacent to the slash can be used as part of the syntax,

a dash "--" or other symbol to indicate that the token is an optional part of the syntax,

a comma "," or other symbol or method to indicate that a token has attached attributes in which the attributes follow the comma.

5. Apparatus according to claim 4, wherein said rule syntax is used to also express the phrase structure rules.

6. A process for the analysis of encoded natural language text comprising the steps of:

(a) reading information of a digitally encoded text and dividing said text into words which can be sequentially examined, starting with the first word in said text

(b) placing the phrase structure rule of a sentence for a language on a stack structure

(c) expanding the phrase structure rule which is on the top of the stack until a token representing a word is reached

(d) checking if the said token matches the word in the sentence that is being examined, skipping to step (h) if no match is found

(e) remove said word token from the stack

(f) retrieve from the dictionary entry for said word a rule which specifies the kinds of words and phrases and their syntactic order which can follow said word and place this rule into the top of the stack

(g) get the next word in said sentence as the word to be examined and return to step (c)

(h) checking if said token must match a word in said text and taking an error recovery action if said token must match a word

(i) remove said non-matching token from the stack

(j) terminate processing if the stack is empty or if there are no more words to be examined in the sentence, otherwise returning to step (c).

7. The process of claim 6, further including the step of enforcing agreement among clauses by transferring attributes into the stack structure.

8. The process of claim 6, further including the step of backtracking and restarting processing if an error is detected in step (h) by saving the state of the stack when an ambiguous word is processed and restoring the saved stack when an error is detected in step (h).