Processing of textual information and automated apprehension of information

• US 6,871,199 B1
• Filed: 05/12/1999
• Issued: 03/22/2005
• Est. Priority Date: 06/02/1998
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bloechl, Peter, Klenk, Juergen, Binnig, Gerd K.
• Primary Examiner(s): Wassum, Luke S
• Application Number:
  US09/310,650
• Time in Patent Office: 2,141 Days
• Field of Search: 704 2- 7, 704/9, 707 3- 5
• US Class Current: 1/1
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

Scheme for the automated apprehension of textual information conveyed in an input string. The input string is segmented to generate segments and/or semantical units. The following steps are repeated for each segment in the input string until a subset for each segment in said input string is identified:

• • a. identifying a matching semantical unit in a fractal hierarchical knowledge database of semantical units and pointers, said matching semantical units being deemed to be related to a segment of said input string,
  • b. determining the fitness of said matching semantical unit by taking into consideration said semantical unit'"'"'s associations,
  • c. defining a subset of information related to said matching semantical unit within said fractal hierarchical knowledge database.

Then these subsets are combined to form a resulting semantic network.

Claims

1. A method of processing textual information, comprising the steps of:

segmenting an input string which includes textual information into segments; and

combining the segments with semantical units from a knowledge database representing a network of hierarchically arranged semantical units which are similar across hierarchies, the combining step generating a resulting semantic network of hierarchically arranged semantical units which are similar across hierarchies.

2. The method of claim 1, wherein at least one of the segments is one of related to a semantical unit and similar to a semantical unit, and wherein there are at least n semantical units in the knowledge database with n being not less than 2.

3. The method of claim 2, wherein a semantical unit is a set containing one or more pieces of information.

4. The method of claim 3, wherein the semantical unit includes a name and one or more pointers to other semantical units.

5. The method of claim 2, wherein the combining step comprises the following steps:

identifying a matching semantical unit in the knowledge database, the matching semantical unit being related to the j^th segment of the input string;

determining a fitness of the matching semantical unit by taking into consideration the possible associations of the semantical unit; and

repeating the identifying and determining steps until a matching semantical unit is found for j=1, . . . , m segments of the input string.

6. The method of claim 5, wherein combining of the segments with the matching semantical units is done-such that the resulting semantic network includes semantical units that are related to the textual information conveyed in the input string.

7. The method of claim 5, whereby identifying and determining steps are carried out m times if there are m matching semantical units in the knowledge database which are related to the j^th semantical unit of the input string, and wherein one of the m matching semantical units is used to generate the resulting semantic network.

8. The method of claim 5, wherein the matching semantical units are combined with each other to form the resulting semantic network if a corresponding classification probability indicates that the respective semantical units have been correctly matched with the semantical units in the knowledge database.

9. The method of claim 5, wherein the possible associations of the semantical unit are at least one of possible attributes, possible relations and possible roles.

10. The method of claim 5, wherein at least one of attributes, relations and roles of a semantical unit in the input string are used to determine whether a matching semantical unit in the knowledge database is better than other matching semantical units in the knowledge database.

11. The method of claim 10, wherein the determination of whether a matching semantical unit in the knowledge database is better than other matching semantical units in the knowledge database includes determining whether a matching semantical unit in the knowledge database has a higher fitness than other matching semantical units in the knowledge database.

12. The method of claim 5, wherein the fitness is increased if at least one of generally accepted attributes, relations and roles are used in connection with a semantical unit.

13. The method of claim 5, wherein the fitness is decreased if at least one of non-generally acceptable attributes, relations and roles are used in connection with a semantical unit.

14. The method of claim 5, wherein the identifying and determining steps are repeated if an additional input string is received until a resulting semantic network is generated.

15. The method of claim 5, wherein one or more self-similar algorithms are used when carrying out the identifying and determining steps.

16. The method of claim 2, wherein the input string is transformed into an input network including the semantical units.

17. The method of claim 16, wherein the resulting semantic network is obtained from transforming the input network.

18. The method of claim 16, wherein the input network is generated by at least one of syntactic parsing and grammatical parsing.

19. The method of claim 1, wherein the resulting semantic network is employed for automated apprehension of the textual information conveyed in the input string.

20. The method of claim 1, whereby the knowledge database and the resulting semantic network include one or more types of semantical units such as objects, relations, and attributes.

21. The method of claim 1, wherein the knowledge database and the resulting semantic network include one or more types of pointers.

22. The method of claim 21, wherein a pointer between two semantical units in the knowledge database has a weight corresponding to a semantical distance between the two semantical units, the weight being one of a fixed weight and a variable weight.

23. The method of claim 22, wherein an inverse of the weight represents a semantical distance between two semantical units which the pointer connects.

24. The method of claim 21, wherein one or more of the pointers describe mutual relationships between semantical units.

25. The method of claim 21, wherein at least one of the pointers is a directed associative connection between semantical units.

26. The method of claim 21, wherein at least one of the pointers is one of a hierarchical pointer, a horizontal pointer, a similarity pointer, a functional pointer, an attributional pointer and a role pointer.

27. The method of claim 1, wherein each semantical unit in the knowledge database receives a potential, the potential being one of a fixed potential and a variable potential.

28. The method of claim 1, wherein the resulting semantic network is a network identified within the knowledge database.

29. The method of claim 1, wherein the resulting semantic network is extracted from the knowledge database to form a new separate network.

30. The method of claim 1, wherein an input text is transformed into one or more input strings.

31. The method of claim 1, wherein a preliminary theme of the input string is determined allowing a quick identification of a preliminary subset of information within the knowledge database.

32. The method of claim 31, wherein the preliminary subset is revised.

33. The method of claim 32, wherein the preliminary subset is revised one of continuously and occasionally.

34. The method of claim 32, wherein the preliminary subset is revised when one of a contradiction and change of theme is determined.

35. The method of claim 1, further comprising the step of performing an action depending on the general meaning extracted from the input string.

36. The method of claim 1, wherein an action is performed by providing an answer if the input string is determined to include a question.

37. The method of claim 1, wherein the knowledge database includes a self-similar representation of semantical units and pointers across different scales.

38. The method of claim 1, wherein the knowledge database reflects the structure of an environment.

39. The method of claim 38, wherein the environment is a subset of the real world.

40. The method of claim 1, wherein text is transformed into input strings which are then processed string-by-string.

41. The method of claim 1, wherein speech is transformed into input strings which are then processed string-by-string.

42. The method of claim 41, wherein speech recognition is employed to transform the speech into said input strings.

43. A method of constructing a fractal hierarchical knowledge database representing a network of hierarchically arranged semantical units which are similar across hierarchies, comprising the steps of:

recording semantical units;

linking the semantical units by one or more pointers with other semantical units of the fractal hierarchical knowledge database; and

assigning a weight to the one or more pointers.

44. The method of claim 43, wherein the one or more pointers are one of hierarchical pointers and horizontal pointers.

45. The method of claim 44, wherein at least one of the hierarchical and horizontal pointers is one of a similarity pointer, a functional pointer, an attributional pointer and a role pointer.

46. The method of claim 43, wherein the one or more pointers have one or more types associated therewith which describe associations between semantical units.

47. The method of claim 43, wherein at least one of the one or more pointers is a semantical unit.

48. The method of claim 43, wherein the weight assigned to the pointer is one of fixed and variable.

49. The method of claim 43, wherein one of a fixed and a variable potential is assigned to a semantical unit.

50. The method of claim 49, wherein the potential is assigned when the fractal hierarchical knowledge database is used for the automated apprehension of textual information conveyed in an input string.

51. The method of claim 50, wherein one of a fixed and a variable threshold is assigned to the fractal hierarchical knowledge database.

52. The method of claim 51, wherein the potential and threshold are used for subset determination.

53. The method of claim 43, wherein an inverse of the weight of the pointer represents a semantic distance between two semantical units which the pointer connects.

54. Apparatus for processing textual information, comprising:

a segmenter which segments an input string which includes textual information into segments;

a semantic processor coupled to the segmenter which combines the segments with semantical units from a knowledge database representing a network of hierarchically arranged semantical units which are similar across hierarchies, the combining operation generating a resulting semantic network of hierarchically arranged semantical units which are similar across hierarchies; and

memory coupled to the semantic processor which stores the segments and the resulting semantic network.

55. The apparatus of claim 54, wherein at least one of the segments is one of related to a semantical unit and identical to a semantical unit, and wherein there are at least n semantical units in the knowledge database with n being not less than 2.

56. The apparatus of claim 55, wherein the semantic processor performs the following operations:

identifying a matching semantical unit in the knowledge database, the matching semantical unit being related to the j^th segment of the input string such that a matching semantical unit is found for j=1, . . . , m segments of the input string; and

determining a fitness of the matching semantical unit by taking into consideration the possible associations of the semantical unit.

57. The apparatus of claim 56, further comprising a speech processing unit which transforms speech into the input string.

58. The apparatus of claim 57, wherein the speech processing unit performs speech recognition.

59. The apparatus of claim 54, wherein the apparatus is used for automated apprehension of the textual information conveyed in the input string.

60. The apparatus of claim 59, wherein a reaction is triggered depending on the apprehended information.

61. The apparatus of claim 54, further comprising a post-processor which transforms the resulting semantic network to create an output string.

62. The apparatus of claim 54, wherein the resulting semantic network is represented as a fractal hierarchical network of semantical units.

63. The apparatus of claim 54, further comprising a processor for processing the input string to generate an input network which includes the semantical units.