Natural language based search engine and methods of use therefor

• US 20060224580A1
• Filed: 04/28/2005
• Published: 10/05/2006
• Est. Priority Date: 03/31/2005
• Status: Active Grant

• Current Assignee: Jiles Incorporated
• Original Assignee: Jiles Incorporated
• Inventors: Quiroga, Martin A., Flowers, John S., DeSanto, John A.
• Granted Patent: US 7,447,683 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9532 Query formulation
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99932 Access augmentation or optimizing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

There is provided a search engine or other electronic search application that receives an inputted query in natural language. The search engine then analyzes the query in accordance with the syntactic relationships of the natural language in which it was presented, weights the syntactic relationships, and generates a result to the query as output, corresponding to the syntactic relationship of the greatest weight. The outputted result is typically an answer, in the form of a sentence or a phrase, along with the document from which the sentence or phrase is taken, including a hypertext link for the document.

Claims

1. A method for providing data comprising:

creating related pairs of words from at least one document in a corpus;

assigning a concept identifier to each word of each related pair of words;

creating pairs of concept identifiers by applying the assigned concept identifiers to each word in each related pair of words;

assigning concept link identifiers to each pair of concept identifiers; and

,

calculating a value for each concept link identifier.

2. The method of claim 1, wherein creating related pairs of words from documents in a corpus, includes:

isolating sentences from the at least one document in a corpus; and

,

parsing the words from each sentence from the at least one document into pairs of words in accordance with predetermined relationships.

3. The method of claim 2, additionally comprising:

creating statements by combining the concept link identifiers from each of the isolated sentences.

4. The method of claim 3, additionally comprising:

storing the statements in at least one data store.

5. The method of claim 4, additionally comprising:

receiving at least one query;

creating related pairs of words in the query;

assigning concept identifiers to the each of the words in each of the related pairs of words;

creating pairs of concept identifiers by applying the assigned concept identifiers to each word in the related pairs of words;

assigning concept link identifiers to each pair of concept identifiers; and

,

combining all of the concept link identifiers into a query statement.

6. The method of claim 5, wherein all of the concept link identifiers of the query statement define a master set, where N is the number of concept link identifiers in the master set;

and,

creating a plurality of subsets from the master set, each subset of the plurality of subsets including at least one member of N concept link identifiers and at least N members of one concept link identifier.

7. The method of claim 6, wherein the creating related pairs of words includes, parsing the query in a parser.

8. The method of claim 6, additionally comprising:

analyzing a plurality of stored statements with the plurality of subsets, including, determining matches of the concept link identifiers in the stored statements with all of the concept link identifiers in each of the subsets, and isolating the stored statements where at least one concept link identifier matches at least one concept link identifier in at least one subset of the plurality of subsets.

9. The method of claim 8, additionally comprising:

calculating the weight of each of the isolated statements.

10. The method of claim 9, wherein calculating the weight of each of the isolated statements, includes, applying a function based on:

the value of the sum of the values for the matching concept link identifiers;

the authority ranking for the document from which the isolated statement was derived; and

the age of the document from which the concept link identifiers were derived, to each of the isolated statements.

11. The method of claim 10, wherein the isolated statement of the greatest weight is determined to be a response to the query.

12. The method of claim 11, wherein the isolated statement of the greatest weight corresponds to at least one sentence of at least one document, and, the at least one sentence is returned to a predetermined location.

13. A method for providing at least one response to at least one query in natural language, comprising:

populating a data store by obtaining documents from at least a portion of a corpus, isolating sentences from the documents, parsing the sentences into linked pairs of words in accordance with predetermined relationships, assigning concept identifiers to each word of the linked pair of words, assigning concept link identifiers to each pair of concept identifiers corresponding to each linked pair of words, and, combining the concept link identifiers for each sentence into a statement;

receiving an inputted query in natural language;

parsing the query into linked pairs of words in accordance with predetermined relationships, assigning concept identifiers to each word of the linked pair of words, assigning concept link identifiers to each pair of concept identifiers corresponding to each linked pair of words, determining values for each of the concept link identifiers, and, combining the concept link identifiers into a query statement;

analyzing the query statement and the statements in the data store for matches between concept link identifiers;

isolating statements in the data store having at least one concept link identifier that matches at least one concept link identifier in the query statement;

calculating the weight of each of the isolated statements from the data store;

providing at least one response to the query, the at least one response including at least one sentence corresponding to the isolated statement from the data store having the greatest weight.

14. The method of claim 13, additionally comprising:

providing access to at least one document from which the at least one sentence was isolated.

15. The method of claim 13, wherein the predetermined relationships are defined by a parser.

16. The method of claim 13, wherein assigning concept identifiers to each word of the query includes, performing a lookup in the data store for the concept identifier matching the word from the query.

17. The method of claim 16, wherein assigning concept link identifiers includes, performing a lookup in the data store for paired concept identifiers matching the paired concept identifiers from the query.

18. A method for providing at least one response to a query comprising:

obtaining at least one document from at least a portion of a corpus, isolating sentences from the at least one document, creating pairs of words from each of the sentences, each pair of words corresponding to a concept link identifier, and creating a statement from at least one concept link identifier from each sentence;

receiving at least one query;

creating a query statement by creating pairs of words from the words that form the at least one query, each pair of words corresponding to a concept link identifier, and the query statement including at least one concept link identifier;

isolating statements having at least one concept link identifier matching at least one concept link identifier of the query statement;

calculating weights for each of the isolated statements; and

,

providing at least one sentence corresponding to the statement of the greatest weight as the at least one response to the at least one query.

19. The method of claim 18, wherein, creating the statements includes storing the statements in a data store.

20. The method of claim 18, wherein calculating the weight (q) for each of the isolated statements includes, calculating a weight (s) for the statement based on the sum of values for the matching concept link identifiers, calculating a value for the authority (av) of the document from which the isolated statement was derived, and calculating a value for the age (d) of the document.

21. The method of claim 20, wherein, the weight for each isolated statement is calculated in accordance with the formula:

q=s·

av/(d+1)

22. The method of claim 21, wherein the authority value (av) is a function of user adjustable weights, x, y and z,

where,

x is a coefficient derived from values corresponding to the author of the document from which the isolated statement was derived,

y is a coefficient derived from values corresponding to the publishing source of the document from which the isolated statement was derived, and,

z is a coefficient derived from values corresponding to the publishing class of the document from which the isolated statement was derived, and,

the authority value (av) is calculated in accordance with the formula:

av=(x+1)·

(y+1)·

(z+1)

where,

0≦

x≦

1,

0≦

y≦

1, and,

0≦

z≦

1.

23. The method of claim 21, wherein the document age (d) is calculated in accordance with the formula:

d=t _now −

h

where,

t_now is the time of the query; and

h is the publishing date of the document.

24. The method of claim 18, wherein the creating the pairs of words from the sentences isolated from the at least one document includes, parsing each sentence in a parser, and linking the words of each pair of words in accordance with predetermined relationships.

25. The method of claim 18, wherein the creating the pairs of words that form the at least one query includes, parsing the at least one query in a parser, and linking the words of each pair of words in accordance with predetermined relationships.

26. An architecture for providing at least one response to a query comprising:

a first module configured for;

obtaining at least one document from at least a portion of a corpus;

isolating sentences from the at least one document;

creating pairs of words from each of the sentences, each pair of words corresponding to a concept link identifier; and

,

creating a statement from at least one concept link identifier from each sentence;

a second module configured for;

receiving at least one query;

creating a query statement by creating pairs of words from the words that form the at least one query, each pair of words corresponding to a concept link identifier, and, the query statement including at least one concept link identifier;

isolating statements having at least one concept link identifier matching at least one concept link identifier of the query statement;

calculating weights for each of the isolated statements; and

,

providing at least one sentence corresponding to the statement of the greatest weight as the at least one response to the at least one query.

27. The architecture of claim 26, additionally comprising:

a data store in communication with the first module and the second module, for storing statements formed from at least one concept link identifier from each sentence.

28. The architecture of claim 26, wherein the second module configured for calculating weights for each of the isolated statements, includes means for:

calculating a weight (s) for the statement based on the sum of values for the matching concept link identifiers, calculating a value for the authority (av) of the document from which the isolated statement was derived, and, calculating a value for the age (d) of the document.

29. The architecture of claim 26, wherein the first module configured for creating the pairs of words from the sentences isolated from the at least one document includes, a parser for parsing each sentence and linking the words of each pair of words from each parsed sentence in accordance with predetermined relationships.

30. The architecture of claim 26, wherein the second module configured for creating the pairs of words that form the at least one query includes, a parser for parsing the at least one query, and linking the words of each pair of words from the parsed at least one query in accordance with predetermined relationships.

31. The architecture of claim 26, wherein the first module additionally includes programmable means through which values corresponding to at least one attribute of a document may be entered.

32. The architecture of claim 31, wherein the at least one attribute of a document includes a plurality of attributes.

33. The architecture of claim 32, wherein the plurality of attributes of a document are selected from the group consisting of:

author, publishing source, and, publishing class.

34. A method for analyzing a query, comprising:

receiving a query in natural language;

obtaining at least one response to the query, the at least one response in accordance with the relationships of the words to each other in natural language of the query; and

,

selecting at least one response to the query having the greatest weight, the weight of the at least one response being a function of the authority of the document from which the response was derived and the time relevance of the relationships of the words to each other in the response.

35. The method of claim 34, wherein obtaining the at least one response to the query includes, providing at least one representation of a corpus, the corpus including text in natural language, based on the relationships of words to each other in natural language.

36. The method of claim 35, wherein obtaining the at least one response to the query includes, creating relational components of the query in accordance with the relationships of the words to each other in the natural language of the query.

37. The method of claim 36, wherein providing the at least one response to the query includes, matching the relational components of the query to a portion of the representation of the corpus.

38. The method of claim 37, wherein the matching includes, isolating at least a portion of the representation of the corpus.

39. The method of claim 38, wherein the at least one response having the greatest weight includes at least one sentence from the text of the corpus corresponding to the isolated portion of the representation of the corpus.

40. A search engine comprising:

a first component configured for receiving a query in natural language; and

,

a second component configured for;

providing at least one response to the query, the at least one response in accordance with the relationships of the words to each other in natural language of the query, and, having the greatest weight, the weight of the at least one response being a function of the authority of the document from which the at least one response was derived and the time relevance of the relationships of the words to each other in the at least one response.

41. The search engine of claim 40, wherein the second component is additionally configured for providing at least one representation of a corpus, the corpus including text in natural language, based on the relationships of words to each other in natural language.

42. The search engine of claim 41, wherein the first component includes a module configured for creating relational components of the query in accordance with the relationships of the words to each other in the natural language of the query.

43. The search engine of claim 42, wherein the module of the first component is additionally configured for matching the relational components of the query to a portion of the at least one representation of corpus.

44. The search engine of claim 43, wherein the first component is additionally configured for isolating the portion of the at least one representation of the corpus that matches the relational components of the query.

45. The search engine of claim 44, wherein the second component is additionally configured for providing at least one sentence from the text of the corpus corresponding to the at least one response having the greatest weight, the at least one response having the greatest weight from the isolated portion of the representation of the corpus.

46. A method of providing responses to a query comprising:

obtaining statements, each statement derived from at least one sentence of at least one document, the at least one document defining at least a portion of a corpus, each of the isolated statements formed from the relationships of the words within the at least one sentence;

isolating at least one of the obtained statements; and

,

calculating a weight for the at least one isolated statement.

47. The method of claim 46, wherein, calculating a weight for the at least one isolated statement includes, calculating the weight as function of the authority of the document from which the statement was derived and the time relevance of the relationships of the words to each other in the at least one isolated statement.

48. The method of claim 47, wherein the relationships of the words to each other in the at least one isolated statement includes pairs of words, each pair of words defining a concept link identifier.

49. The method of claim 48, wherein the weight (q) is calculated in accordance with the formula:

q=s·

av/(d+1)

where,

s is a weight for the statement based on the sum of values for a predetermined number of concept links in the statement;

av is the authority value of the document from which the statement was derived, and,

d is the age of the document.

50. An architecture for providing responses to a query comprising:

a first component configured for obtaining statements, each statement derived from at least one sentence of at least one document, the at least one document defining at least a portion of a corpus, each of the isolated statements formed from the relationships of the words within the at least one sentence;

a second component configured for isolating at least one of the obtained statements; and

,

a third component configured for calculating a weight for the at least one isolated statement.

51. The architecture of claim 50, wherein, the third component is additionally configured for calculating a weight for the at least one isolated statement as function of the authority of the document from which the statement was derived and the time relevance of the relationships of the words to each other in the at least one isolated statement.

52. The architecture of claim 51, wherein the relationships of the words to each other in the at least one isolated statement include, pairs of words, each pair of words defining a concept link identifier, and the third component is additionally configured for calculating the weight (q) in accordance with the formula:

q=s·

av/(d+1)

where,

s is a weight for the statement based on the sum of values for a predetermined number of concept links in the statement;

av is the authority value of the document from which the statement was derived, and,

d is the age of the document.

53. A programmable storage device readable by a machine, tangibly embodying a program of instructions executable by a machine to perform method steps for providing responses to a query, the method steps selectively executed during the time when the program of instructions is executed on the machine comprising:

obtaining statements, each statement derived from at least one sentence of at least one document, the at least one document defining at least a portion of a corpus, each of the isolated statements formed from the relationships of the words within the at least one sentence;

isolating at least one of the obtained statements; and

,

calculating a weight for the at least one isolated statement.

54. The programmable storage device of claim 53, wherein the calculating a weight for the at least one isolated statement includes:

calculating a weight for the at least one isolated statement as function of the authority of the document from which the statement was derived and the time relevance of the relationships of the words to each other in the at least one isolated statement.

55. The programmable storage device of claim 53, wherein the relationships of the words to each other in the at least one isolated statement includes pairs of words, each pair of words defining a concept link identifier, and, calculating the weight (q) in accordance with the formula:

q=s·

av/(d+1)

where,

s is a weight for the statement based on the sum of values for a predetermined number of concept links in the statement;

av is the authority value of the document from which the statement was derived, and,

d is the age of the document.

56. A programmable storage device readable by a machine, tangibly embodying a program of instructions executable by a machine to perform method steps for providing responses to a query, the method steps selectively executed during the time when the program of instructions is executed on the machine comprising:

receiving a query in natural language;

obtaining at least one response to the query, the at least one response in accordance with the relationships of the words to each other in natural language of the query; and

,

selecting at least one response to the query having the greatest weight, the weight of the at least one response being a function of the authority of the document from which the response was derived and the time relevance of the relationships of the words to each other in the response.

57. The programmable storage device of claim 56, wherein obtaining the at least one response to the query includes, creating relational components of the query in accordance with the relationships of the words to each other in the natural language of the query.

58. The programmable storage device of claim 57, wherein providing the at least one response to the query includes, matching the relational components of the query to the relational components in a representation of a corpus, the corpus including text in natural language, and the representation of the corpus including the relational components based on relationships of words to each other in the text in natural language.

59. The programmable storage device of claim 58, wherein the matching includes, isolating the portion of the representation of the corpus with at least one matching relational component.