Method and apparatus for using directed reasoning to respond to natural language queries

• US 8,719,005 B1
• Filed: 02/09/2007
• Issued: 05/06/2014
• Est. Priority Date: 02/10/2006
• Status: Active Grant

• Current Assignee: Marc Joseph Angelone, Rusty Shawn Lee
• Original Assignee: Marc Joseph Angelone, Rusty Shawn Lee
• Inventors: Lee, Rusty Shawn, Angelone, Marc Joseph
• Primary Examiner(s): PULLIAS, JESSE SCOTT
• Application Number:
  US11/672,969
• Time in Patent Office: 2,643 Days
• Field of Search: 707/706, 707/708, 707/755, 707/771
• US Class Current: 704/9
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

A method and apparatus are presented for using directed reasoning to provide natural language responses to questions and commands in natural language. The directed reasoning process uses concepts in the natural language sentence to identify relevant response concepts in a knowledge base, by creating a continuous chain of logic from the input concept to the response concept and limiting the creation of the logic chains to only those consistent with the context of the natural language sentence and information in the knowledge base. The invention allows the identification of relevant concepts that elude traditional concept searching based on pattern matching, while focusing computational resources on only those logic calculations relevant to the natural language sentence and knowledge base. The invention will automatically identify the relevant source documents to use and create a knowledge base during run time, without requiring preprogrammed ontologies or knowledge bases.

Claims

1. A computer implemented method for responding to a natural language sentence in which:

the terms;

“

reasoning”

is;

using a first concept in a natural language sentence to identify a second concept by creating a continuous chain of logic from the first concept to the second concept; and

“

directed reasoning”

is;

applying reasoning to a first concept in a natural language sentence to identify a second concept, wherein only reasoning related to the natural language sentence is allowed,

the method comprising the steps of;

receiving, in a computer, at least one input concept, wherein the at least one input concept is generated from parsing an initial natural language sentence;

searching a knowledge base comprising a plurality of natural language sentences using the at least one input concept to retrieve at least one response concept from the knowledge base, the searching step further comprising the step of resolving one or more contextual referents occurring in a natural language sentence of the plurality, wherein each of the one or more contextual referents is a word within the natural language sentence, the word referring to a previously parsed word or a phrase in the natural language sentence or another natural language sentence of the plurality;

performing directed reasoning on both the at least one input concept and the at least one response concept to select at least one selected response concept by connecting the at least one response concept to the at least one input concept through a continuous chain of logic based on at least one of inductive reasoning, deductive reasoning or abductive reasoning;

and

generating at least one response sentence containing the at least one selected response concept.

2. The method of claim 1, further comprising the step of outputting at least one response sentence containing the selected response concept, wherein the initial natural language sentence is a question and the at least one response sentence is an answer to the question.

3. The method of claim 1, wherein the knowledge base comprises at least one document derived from the World Wide Web.

4. The method of claim 1, further comprising the step of parsing the initial natural language sentence into at least one syntax tree.

5. The method of claim 1, further comprising the step of parsing the initial natural language sentence into at least one syntax tree, wherein the parsing step includes using directed reasoning to eliminate illogical or nonsensical parses.

6. The method of claim 1, wherein the knowledge base comprises pre-existing natural language sentences found in an information source.

7. The method of claim 6, further comprising the steps of:

constructing new natural language sentences from the pre-existing natural language sentences, wherein the new natural language sentences are not found in the information source; and

adding the new natural language sentences to the knowledge base.

8. The method of claim 1, further comprising the step of using directed reasoning to select relevant contextual advertising.

9. The method of claim 1, wherein the resolving step further comprises the step of categorizing the contextual referent into a category.

10. The method of claim 9, wherein the category is selected from the group consisting of:

noun phrase, action, masculine noun phrase, feminine noun phrase, neuter noun phrase, verb phrase, and sentence fragment.

11. The method of claim 9, by substituting a reference for the most recent occurrence in the category that does not create a contradiction or violate constraints in the knowledge base.

12. A computer implemented data processing system for responding to a natural language sentence, comprising:

a server computer having computer readable program code means, the computer readable program code means further comprising;

means for obtaining at least one input concept, wherein the at least one input concept is generated from parsing an initial natural language sentence;

means for searching a knowledge base using the at least one input concept to retrieve at least one response concept from the knowledge base, and resolving a contextual referent occurring in a natural language sentence of the knowledge base by categorizing the contextual referent into a category and substituting a reference for the most recent occurrence in the category that does not create a contradiction or violate constraints in the knowledge base;

means for performing directed reasoning on the at least one input concept and the at least one response concept and selecting at least one of the response concept by connecting the response concept to the input concept through a continuous chain of logic based on at least one of inductive reasoning, deductive reasoning or abductive reasoning;

and

means for generating at least one response sentence containing the at least one selected concept.

13. The system of claim 12, wherein the initial natural language sentence is a command that executes the effects described by the natural language sentence.

14. The system of claim 12, wherein the knowledge base comprises pre-existing natural language sentences found in an information source.

15. The system of claim 14, further comprising:

means for constructing new natural language sentences from the pre-existing natural language sentences, wherein the new natural language sentences are not found in the information source; and

means for adding the new natural language sentences to the knowledge base.

16. The method of claim 12, wherein the category is selected from the group consisting of:

noun phrase, action, masculine noun phrase, feminine noun phrase, neuter noun phrase, verb phrase, and sentence fragment.

17. A computer-implemented method of responding to a natural language sentence, comprising:

parsing an initial natural language sentence to obtain at least one search term;

searching, on a computer, an information source using the at least one search term and retrieving at least one pre-existing natural language text relating to the search term;

reading the at least one pre-existing natural language text and storing the at least one pre-existing natural language text in a knowledge base, the reading step including the step of applying directed reasoning to resolve one or more contextual referents within the at least one pre-existing natural language text, wherein each of the one or more contextual referents is a word that refers to a previously parsed word or a previously parsed phrase in the at least one pre-existing natural language text;

constructing at least one new natural language text from the at least one preexisting natural language text;

generating at least one response derived from the at least one new natural language text.

18. The method of claim 17, further comprising the step of outputting the at least one response, wherein the initial natural language sentence is a question and the at least one response is an answer to the question.

19. The method of claim 17, wherein the information source is the World Wide Web.

20. The method of claim 17, wherein the parsing step further comprising the step of using the initial natural language sentence to prevent or eliminate a search term that is inconsistent with or irrelevant to the initial natural language sentence.

21. The method of claim 17, wherein the searching step further comprising the step of using the initial natural language sentence to prevent or eliminate the retrieval of textual data that is inconsistent with or irrelevant to the initial natural language sentence.

22. The method of claim 17, wherein the constructing step further comprising the step of using the initial natural language sentence to prevent or eliminate the construction of a new natural language text that is inconsistent with or irrelevant to the initial natural language sentence.

23. The method of claim 22, wherein the new natural language text is constructed from the pre-existing natural language texts by applying a logical reasoning process using at least one of inductive reasoning, deductive reasoning or abductive reasoning.

24. The method of claim 17, wherein the parsing step further comprises the step of resolving a contextual referent occurring in the initial natural language sentence by categorizing the contextual referent into a category and substituting a reference for the most recent occurrence in the category that does not create a contradiction or violate constraints in a knowledge base, the category selected from the group consisting of:

noun phrase, action, masculine noun phrase, feminine noun phrase, neuter noun phrase, verb phrase, and sentence fragment.