Control signal processing method and apparatus having natural language interfacing capabilities

• US 5,870,701 A
• Filed: 06/09/1997
• Issued: 02/09/1999
• Est. Priority Date: 08/21/1992
• Status: Expired due to Term

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Wachtel, Thomas Juliusz
• Primary Examiner(s): Thomas, Joseph
• Application Number:
  US08/871,846
• Time in Patent Office: 610 Days
• Field of Search: 704/1, 704/9, 704/10, 704/257, 704/243, 345/326, 345/968, 345/969, 364/146, 364/188, 395/12
• US Class Current: 704/9
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching

Abstract

A signal processing method and apparatus suitable for processing input signals representative, for example, of words or other elemental units of a natural language, which signals fall into a plurality of different categories, is provided in which the category of each input signal is determined, search data dependent upon the category is stored, the search data stored in response to previous signals is searched in response to each new signal, the search performed being dependent upon the category of the new signal, and an output signal dependent upon the input signals and upon the results of the searches is provided. When natural language is processed, the output signals may thus be dependent upon the meaning of groups of input words or other elemental units and is provided in a form in which it may conveniently be further processed. An application of the invention in a computer control switching system is described enabling the operator to input switching instructions in natural language.

Claims

1. A natural language interface to a computer which generates data structured in computer usable form and derived from input data in the form of word units in the natural language, said word units comprised of a plurality of different lexical categories, said interface comprising:

(a) first defining means for defining the word units of said natural language which may be processed by the interface;

(b) data unit generating means for generating computer usable data units in response to input of said word units of said natural language, wherein different data units are generated for different word units of said natural language;

(c) second defining means for defining a plurality of different search codes;

(d) selecting means for selecting, in response to input of word units of at least some of said lexical categories, search codes from said plurality of different search codes, the selected search codes being selected based on the lexical category of the input word unit, the selected search codes being usable to generate links only between the data unit generated by said generating means in response to input of the current word unit and one or more subsequently generated data units;

(e) searching means for searching, in response to input of word units in at least some of said lexical categories, for predetermined search codes which have been previously selected, the predetermined search codes being dependent upon the lexical category of the word unit; and

(f) link generating means responsive to said predetermined search codes being found by said searching means for generating links between data units to provide said data structured in computer usable form.

2. A natural language interface according to claim 1, wherein said data unit generating means comprises a lexicon and means for accessing said lexicon to obtain therefrom both the lexical category of the word units and said data units.

3. A natural language interface according to claim 1, wherein said selecting means is operable to select exactly one of said search codes in response to input word units of at least one of said lexical categories.

4. A natural language interface according to claim 1, wherein said selecting means is operable to select a plurality of different search codes in response to input word units of at least one of said lexical categories.

5. A natural language interface according to claim 1, wherein said searching means is operable to perform no search for said predetermined search codes in response to input word units of at least one of said lexical categories.

6. A natural language interface according to claim 1, wherein said searching means is operable to perform a search for exactly one of said predetermined search codes in response to input word units of at least one of said lexical categories.

7. A natural language interface according to claim 1, wherein said searching means is operable to perform a search for a plurality of different predetermined search codes in response to input word units of at least one of said lexical categories.

8. A natural language interface according to claim 1, wherein said selecting means and said searching means comprise a look-up table containing instructions for said selecting and for said searching for each of said lexical categories.

9. A natural language interface according to claim 1, wherein, for at least one of said lexical categories, a predetermined search code is deselected if found in a search performed in response to word units of said at least one category.

10. A natural language interface according to claim 9, wherein said selecting means and said searching means comprise a look-up table containing instructions for said selecting and for said searching for each of said lexical categories, and wherein said look-up table includes instructions for performing said deselection.

11. A natural language interface according to claim 1, wherein each said data unit comprises a first part identifying said data unit and a second part for defining said links, said link generating means being operable to generate said links utilizing second parts of said data units.

12. A natural language interface according to claim 11, wherein said link generating means is operable to assign values to the second parts of said data units, and is also operable to generate said links by relating different values of different data units to each other dependent upon searching results of said searching means.

13. A natural language interface according to claim 11, wherein each said data unit is in the form of a predicate having an argument, the first part of said data unit comprising the predicate and the second part comprising the argument.

14. A natural language interface according to claim 13, wherein said link generating means is operable to assign successive values in a series to said arguments as successive said data units are generated, and is also operable to generate said links by equating values assigned to arguments of different data units to each other in response to searching results of said searching means.

15. A natural language interface according to claim 14, wherein said link generating means is operable to instantiate values in said arguments dependent upon the equated values formed.

16. A natural language interface according to claim 13, wherein each said data unit corresponding to a word unit in lexical categories in a first group is provided with at least two arguments, and wherein each said data unit corresponding to a word unit in lexical categories in a second group is provided with at least three arguments.

17. A natural language interface according to claim 16, wherein said lexical categories in said first group include at least one category from the following:

nouns, auxiliary verbs, negation, "to" as an infinitival particle, intransitive verbs, determiners, adjectives, adverbs and intensifiers.

18. A natural language interface according to claim 16, wherein said lexical categories of said second group comprise at least one category from the following:

prepositions, transitive verbs and "to verbs".

19. A natural language interface according to claim 16, wherein said link generating means, when generating links in response to a current data unit generated from a word unit of a lexical category of said first group, (a) utilizes a first of said arguments for generating links between the current data unit and another data unit generated from a word unit which qualifies the word unit from which said current data unit is generated, and (b) utilizes a second of said arguments for generating links between the current data unit and at least one other data unit generated from a word unit which is qualified by the word unit from which the current data unit is generated.

20. A natural language interface according to claim 16, wherein said link generating means, when generating links in response to a current data unit generated from a word unit of a lexical category of said second group, (a) utilizes a first of said arguments for generating links between the current data unit and another data unit generated from a word unit which qualifies the word unit from which said current data unit is generated, and (b) utilizes said second and third arguments for generating links between the current data unit and at least one other data unit generated from a word unit which is qualified by the word unit from which the current data unit is generated.

21. A natural language interface according to claim 13, wherein said link generating means is operable, when generating links in response to a data unit and in response to a predetermined search result of said searching means, to generate a plurality of links between the current data unit and a plurality of previously generated said data units, wherein said plurality of links are defined utilizing the same argument of the current data unit, and wherein said link generating means is further operable to generate a different respective link for each of said plurality of defined links.

22. A natural language interface according to claim 13, wherein said link generating means is operable, in response to a predetermined search result of said searching means, to generate a link between each of a plurality of said data units on the one hand, and a common previously generated data unit on the other hand, utilizing the same argument of said common previously generated data unit for defining each of said links, and wherein link generating means is operable to define a different link corresponding to each of said plurality of links.

23. A natural language interface according to any one of claims 13 to 22, wherein said link generating means is operable to generate first and second links between a first data unit and a second data unit, one of said first and second links being generated utilizing one argument of said first data unit and one of plural arguments of said second data units, and the other of said first and second links being defined by the same argument of said first data unit and a different argument of said second data unit, and said link generating means is operable to define different links for each of said links.

24. A natural language interface according to claim 1, wherein said searching means is operable, when searching in response to a data unit generated in response to input of a word unit which may qualify a previous word unit, to search for all occurrences of the predetermined search codes which may have been selected in response to a data unit previously generated from said previous word unit, and wherein said link generating means is operable to generate a link for each occurrence found.

25. A natural language interface according to claim 1, wherein said searching means is operable, when searching in response to a data unit generated in response to a word unit which may be qualified by a previous word unit, to terminate said search upon finding the latest selected occurrence of a search code selected in response to a data unit generated from said previous word unit, and wherein said link generating means is operable to generate a link for each occurrence found.

26. A natural language interface according to claim 1, wherein said selecting means selects search codes subsequent to said searching means searching for predetermined search codes.

27. A natural language interface according to claim 1, wherein said data unit generating means is operable to proceed with the generation of each data unit without awaiting receipt of any subsequent data unit.

28. A natural language interface according to claim 1, including control signal generating means responsive to said data structured in computer usable form for generating a control signal dependent thereon.

29. A natural language interface according to claim 28, wherein said control signal generating means includes a knowledge base and means for processing said data structured in computer usable form against said knowledge base for generating said control signal.

61. Switching apparatus comprising a plurality of switches and computer apparatus according to claim 28 for controlling said switches, said computer apparatus being arranged for receiving instructions for actuating said switches, for generating from said instructions said data units, and for generating control signals so as to actuate said switches in response to said instructions.

30. A method for providing a natural language interface to a computer, the interface generating data structured in a computer usable form derived from input data in the form of word units in the natural language, said word units comprised of a plurality of different lexical categories, said method comprising the steps of:

(a) defining the word units of said natural language which may be processed by the interface;

(b) generating computer usable data units in response to input of said word units of said natural language, wherein different data units are generated for different word units of said natural language;

(c) defining a plurality of different search codes;

(d) selecting, in response to input of word units of at least some of said lexical categories, search codes from said plurality of different search codes, the selected search codes being selected based on the lexical category of the input word unit, the selected search codes being usable to generate links only between the data unit generated in said generating step in response to input of the current word unit and one or more subsequently generated data units;

(e) searching, in response to input of word units in at least some of said lexical categories, for predetermined search codes which have been previously selected, the predetermined search codes being dependent upon the lexical category of the word unit; and

(f) in response to said predetermined search codes being found in said searching step, generating links between data units to provide said data structured in computer usable form.

31. A method according to claim 30, wherein said step of generating data units comprises accessing a lexicon to obtain therefrom both the lexical category of the word units and said data units.

32. A method according to claim 30, wherein in said selecting step exactly one of said search codes is selected in response to input word units of at least one of said lexical categories.

33. A method according to claim 30, wherein in said selecting step a plurality of different search codes are selected in response to input word units of at least one of said lexical categories.

34. A method according to claim 30, wherein in said searching step no search for said predetermined search codes is performed in response to input word units of at least one of said lexical categories.

35. A method according to claim 30, wherein in said searching step a search for exactly one of said predetermined search codes is performed in response to input word units of at least one of said lexical categories.

36. A method according to claim 30, wherein in said searching step a search for a plurality of different predetermined search codes is performed in response to input word units of at least one of said lexical categories.

37. A method according to claim 30, wherein said selecting step and said searching step are performed in accordance with instructions stored in a look-up table.

38. A method according to claim 30, wherein, for at least one of said lexical categories, a predetermined search code is deselected if found in a search performed in response to word units of said at least one category.

39. A method according to claim 38, wherein said selecting step and said searching step are performed in accordance with instructions stored in a look-up table, and wherein said look-up table includes instructions for performing said deselection.

40. A method according to claim 30, wherein each said data unit comprises a first part identifying said data unit and a second part for defining said links, said link generating step being performed so as to generate said links utilizing second parts of said data units.

41. A method according to claim 40, wherein said link generating step is performed so as to assign values to the second parts of said data units, and so as to generate said links by relating different values of different data units to each other dependent upon searching results of said searching step.

42. A method according to claim 40, wherein each said data unit is in the form of a predicate having an argument, the first part of said data unit comprising the predicate and the second part comprising the argument.

43. A method according to claim 42, wherein said link generating step is performed so as to assign successive values in a series to said arguments as successive said data units are generated, and so as to generate said links by equating values assigned to arguments of different data units to each other in response to searching results of said searching step.

44. A method according to claim 43, wherein said link generating step is performed so as to instantiate values in said arguments dependent upon the equated values formed.

45. A method according to claim 42, wherein each said data unit corresponding to a word unit in lexical categories in a first group is provided with at least two arguments, and wherein each said data unit corresponding to a word unit in lexical categories in a second group is provided with at least three arguments.

46. A method according to claim 45, wherein said lexical categories in said first group include at least one category from the following:

nouns, auxiliary verbs, negation, "to" as an infinitival particle, intransitive verbs, determiners, adjectives, adverbs and intensifiers.

47. A method according to claim 45, wherein said lexical categories of said second group comprise at least one category from the following:

prepositions, transitive verbs and "to verbs".

48. A method according to claim 45, wherein said link generating step, when generating links in response to a current data unit generated from a word unit of a lexical category of said first group, (a) utilizes a first of said arguments to generate links between the current data unit and another data unit generated from a word unit which qualifies the word unit from which said current data unit is generated, and (b) utilizes a second of said arguments to generate links between the current data unit and at least one other data unit generated from a word unit which is qualified by the word unit from which the current data unit is generated.

49. A method according to claim 45, wherein in said link generating step, when generating links in response to a current data unit generated from a word unit of a lexical category of said second group, (a) utilizes a first of said arguments to generate links between the current data unit and another data unit generated from a word unit which qualifies the word unit from which said current data unit is generated, and (b) utilizes said second and third arguments to generate links between the current data unit and at least one other data unit generated from a word unit which is qualified by the word unit from which the current data unit is generated.

50. A method according to claim 42, wherein in said link generating step, when generating links in response to a data unit and in response to a predetermined search result of said searching step, generates a plurality of links between the current data unit and a plurality of previously generated data units, wherein said plurality of links are defined utilizing the same argument of the current data unit, and wherein said link generating step further generates a different respective link for each of said plurality of defined links.

51. A method according to claim 42, wherein in said link generating step, in response to a predetermined search result of said searching step, a link is generated between each of a plurality of said data units on the one hand, and a common previously generated data unit on the other hand, utilizing the same argument of said common previously generated data unit for defining each of said links, and wherein a different link is defined corresponding to each of said plurality of links.

52. A method according to any one of claims 42 to 51, wherein in said link generating step, first and second links are generated between a first data unit and a second data unit, one of said first and second links being generated utilizing one argument of said first data unit and one of plural arguments of said second data units, and the other of said first and second links being defined by the same argument of said first data unit and a different argument of said second data unit, and different links are defined for each of said links.

53. A method according to claim 30, wherein in said searching step, when searching in response to a data unit generated in response to input of a word unit which may qualify a previous word unit, all occurrences of the predetermined search codes are searched for which may have been selected in response to a data unit previously generated from said previous word unit, and wherein said link is generated for each occurrence found.

54. A method according to claim 30, wherein in said searching step, when searching in response to a data unit generated in response to a word unit which may be qualified by a previous word unit, said search is terminated upon finding the latest selected occurrence of a search code selected in response to a data unit generated from said previous word unit, and wherein said link is generated for each occurrence found.

55. A method according to claim 30, wherein said selecting step selects search codes subsequent to said searching step searching for predetermined search codes.

56. A method according to claim 30, wherein in said data unit generating step, the generation of each data unit is proceeded with without awaiting receipt of any subsequent data unit.

57. A method according to claim 30, further comprising a control signal generating step in which a control signal is generated dependent on said data structured in computer usable form.

58. A method according to claim 57, wherein in said control signal generating step includes processing said data structured in computer usable form against said knowledge base for generating said control signal.

62. A switching method for switching a plurality of switches including a method according to claim 57 for controlling said switches in which instructions are received for actuating said switches, said data units are generated from said instructions, and said control signals generated so as to actuate said switches in response to said instructions.

59. A record carrier storing program instructions to cause a computer to provide an interface which generates data structured in a computer usable form derived from input data in the form of word units in a natural language, said word units comprised of a plurality of lexical different categories, said stored program instructions comprising:

(a) instructions to cause the computer to define the word units of said natural language which may be processed by the interface;

(b) instructions to cause the computer to generate computer usable data units in response to input of said word units of said natural language, wherein different data units are generated for different word units of said natural language;

(c) instructions to cause the computer to define a plurality of different search codes;

(d) instructions to cause the computer to select, in response to input of word units of at least some of said lexical categories, search codes from said plurality of different search codes, the selected search codes being selected based on the lexical category of the input word unit, the selected search codes being usable to generate links only between the data unit generated in response to input of the current word unit and one or more subsequently generated data units;

(e) instructions to cause the computer to search, in response to input of word units in at least some of said lexical categories, for predetermined search codes which have been previously selected, the predetermined search codes being dependent upon the lexical category of the word unit; and

(f) instructions to cause the computer to respond to said predetermined search codes being found and to generate links between data units to provide said data structured in computer usable form.

60. Computer-executable process steps, stored on a computer-readable medium, which provide a natural language interface to a computer by generating data structured in a computer usable form derived from input data in the form of word units in the natural language, said computer-executable process steps comprising:

(a) a step to define the word units of said natural language which may be processed by the interface;

(b) a step to generate computer usable data units in response to input of said word units of said natural language, wherein different data units are generated for different word units of said natural language;

(c) a step to define a plurality of different search codes;

(d) a step to select, in response to input of word units of at least some of said lexical categories, search codes from said plurality of different search codes, the selected search codes being selected based on the lexical category of the input word unit, the selected search codes being usable to generate links only between the data unit generated in said generating step in response to input of the current word unit and one or more subsequently generated data units;

(e) a step to search, in response to input of word units in at least some of said lexical categories, for predetermined search codes which have been previously selected, the predetermined search codes being dependent upon the lexical category of the word unit; and

(f) a step to generate links between data units to provide said data structured in computer usable form, in response to said predetermined search codes found.