Methods and apparatus for thematic parsing of discourse

• US 5,708,822 A
• Filed: 05/31/1995
• Issued: 01/13/1998
• Est. Priority Date: 05/31/1995
• Status: Expired due to Term

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle Corporation
• Inventors: Wical, Kelly
• Primary Examiner(s): Hofsass, Jeffery
• Assistant Examiner(s): HILL, ANDREW M
• Application Number:
  US08/454,602
• Time in Patent Office: 958 Days
• Field of Search: 395/751, 395/759, 395/760, 395/793, 395/794
• US Class Current: 704/1
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique

Abstract

A theme parsing system that determines the thematic context of input discourse is disclosed. Each word used in a language carries thematic information that conveys the importance of the meaning and content of the discourse. The theme parsing system discriminates words and phrases of the input discourse, identifying the type of importance or meaning, the impact on different parts of the discourse, and the overall contribution to the content of the discourse. The thematic conext of the discourse is determined in accordance with predetermined theme assessment criteria that is a function of the strategic importance of the discriminated words. The predetermined thematic assessment criteria defines which of the discriminated words are to be selected for each thematic analysis unit. The discourse is then output in a predetermined thematic format as a different view to a user, giving the topics of the discourse in a topic extractor, generating summarized versions of the discourse in a kernel generator, and identifying the key content of the discourse in a content extractor.

Claims

1. A method for parsing thematic context of input discourse, said method comprising the steps of:

specifying a plurality of thematic constructions that define content of discourse;

testing each of said thematic constructions with said input discourse to identify applicability of each of said thematic constructions to said input discourse; and

generating a thematic context output for said input discourse by generating, for each word in said input discourse, a thematic tag to indicate applicability of each of said thematic constructions to said input discourse.

2. The method as set forth in claim 1, wherein the step of testing each of said thematic constructions with said input discourse comprises the steps of:

receiving contextual tags for said input discourse that identify contextual relationships for words in said input discourse; and

utilizing said contextual tags to determine the applicability of each of said thematic constructions to said input discourse.

3. The method as set forth in claim 1, further comprising the steps of:

referencing, in combination, said thematic tags; and

generating additional thematic tags based on a pre-defined relationship of said thematic tags.

4. The method as set forth in claim 1, further comprising the steps of:

referencing, in combination, pre-defined combinations of said thematic tags from said input discourse;

determining a location of said pre-defined combinations of said thematic tags in said input discourse; and

generating a pointer to identify said location of said pre-defined combinations of said thematic tags in said input discourse.

5. The method as set forth in claim 1, wherein the step of testing each of said thematic constructions with said input discourse comprises the steps of:

storing, in a lexicon, a plurality of words and a plurality of flags for each word to indicate a plurality of definitional characteristics for each word;

accessing said lexicon to read said flags for at least one word; and

utilizing said flags to determine the applicability of each of said thematic constructions to said input discourse.

6. The method as set forth in claim 5, further comprising the step of storing in said lexicon a noun thematic aspect for a plurality of nouns that characterizes a content carrying strength for each noun stored.

7. The method as set forth in claim 3, further comprising the steps of:

storing, in said lexicon, industry specific terminology;

comparing each word in said input discourse with said lexicon to determine whether said word is an industry specific term; and

generating a thematic tag for a word to indicate that said word requires supporting information when said word is an industry specific term.

8. A method for determining thematic context of input discourse, said method comprising the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions to thematic units of said input discourse; and

searching said thematic context output for pre-defined combinations of said thematic tags; and

extracting, in combination, said thematic units corresponding to said pre-defined combination of said thematic tags to extract a theme for said input stream.

9. A method for generating kernel sentences from input discourse, said method comprising the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions including a plurality of thematic contextual elements for said input discourse; and

accessing said thematic tags in said thematic context output; and

removing words from each sentence of said input discourse if a thematic tag for a corresponding word indicates applicability to one of said thematic contextual elements, wherein words that exist define a kernel sentence.

10. A method for determining a topic for each sentence in an input discourse, said method comprising the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions that define content for said input discourse;

accessing said thematic tags in said thematic context output; and

extracting a main topic for said sentence based on said thematic tags including extracting context words that support said main topic.

11. The method as set forth in claim 10, further comprising the step of extracting a secondary topic for said sentence including extracting supporting context words for said secondary topic based on said thematic tags.

12. The method as set forth in claim 10, wherein the step of extracting words that define said main topic of said sentence comprises the step of:

determining, from said thematic tags, whether a first set of thematic assessments exist, said first set of thematic assessments identifying words not used in said main topic of said sentence;

removing words from said sentence that are applicable to said first set of thematic assessments;

determining, from said thematic tags, whether a second set of thematic assessments exist, said second set of thematic assessments identify words that include contextual elements remaining in said sentence but that are not essential to conveying said main topic; and

removing words from said sentence that are applicable to said second set of thematic assessments, wherein words that exist define said main topic of said sentence.

13. A method for extracting content in an input discourse, said method comprising the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions that define content for said input discourse; and

extracting, for each sentence in said input discourse, words that identify at least one of a plurality of major thematic points based on said thematic tags, wherein said words extracted identify content for each sentence in said input discourse.

14. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise thematic head words.

15. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise proper names.

16. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise descriptions.

17. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise named topics.

18. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise topic indicators.

19. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise definitions.

20. The method as set forth in claim 13, wherein at least one of a plurality of major thematic points comprise developed proper names.

21. A computer readable medium for storing instructions, which when executed by a computer, causes the computer to perform the steps of:

storing a plurality of thematic constructions that define content of discourse;

testing each of said thematic constructions with said input discourse to identify applicability of each of said thematic constructions to said input discourse; and

generating a thematic context output for said input discourse by generating, for each word in said input discourse, a thematic tag to indicate applicability of each of said thematic constructions to said input discourse.

22. The computer readable medium as set forth in claim 21, wherein the instructions for testing each of said thematic constructions with said input discourse comprise instructions for:

receiving contextual tags for said input discourse that identify contextual relationships words in said input discourse; and

utilizing said contextual tags to determine the applicability of each of said thematic constructions to said input discourse.

23. The computer readable medium as set forth in claim 21, further comprising instructions for:

referencing, in combination, said thematic tags; and

generating additional thematic tags based on a pre-defined relationship of said thematic tags.

24. The computer readable medium as set forth in claim 21, further comprising instructions for:

referencing, in combination, pre-defined combinations of said thematic tags from said input discourse;

determining a location of said pre-defined combinations of said thematic tags in said input discourse; and

generating a pointer to identify said location of said pre-defined combinations of said thematic tags in said input discourse.

25. The computer readable medium as set forth in claim 21, wherein instructions for testing each of said thematic constructions with said input discourse comprise instructions for:

storing a lexicon comprising a plurality of words and a plurality of flags for each word to indicate a plurality of definitional characteristics for each word;

accessing said lexicon to read said flags for at least one word; and

utilizing said flags to determine the applicability of each of said thematic constructions to said input discourse.

26. The computer readable medium as set forth in claim 25, further comprising instructions for storing in said lexicon a noun thematic aspect for a plurality of nouns that characterizes a content carrying strength for each noun stored.

27. The computer readable medium as set forth in claim 23, further comprising instructions for:

storing, in said lexicon, industry specific terminology;

comparing each word in said input discourse with said lexicon to determine whether said word is an industry specific term; and

generating a thematic tag for a word to indicate that said word requires supporting information when said word is an industry specific term.

28. A computer readable medium for storing instructions, which when executed by a computer, causes the computer to perform the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions to thematic units of said input discourse; and

searching said thematic context output for pre-defined combinations of said thematic tags; and

extracting, in combination, said thematic units corresponding to said pre-defined combination of said thematic tags to extract a theme for said input stream.

29. A computer readable medium for storing instructions, which when executed by a computer, causes the computer to perform the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions including a plurality of thematic contextual elements for said input discourse; and

accessing said thematic tags in said thematic context output; and

removing words from each sentence of said input discourse if a thematic tag for a corresponding word indicates applicability to one of said thematic contextual elements, wherein words that exist define a kernel sentence.

30. A computer readable medium for storing instructions, which when executed by a computer, causes the computer to perform the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions that define content for said input discourse; and

accessing said thematic tags in said thematic context output; and

extracting a main topic for said sentence based on said thematic tags including extracting context words that support said main topic.

31. The computer readable medium as set forth in claim 30, further comprising instructions for extracting a secondary topic for said sentence including extracting supporting context words for said secondary topic based on said thematic tags.

32. The computer readable medium as set forth in claim 30, wherein the step of extracting words that define said main topic of said sentence comprises the step of:

determining, from said thematic tags, whether a first set of thematic assessments exist, said first set of thematic assessments identifying words not used in said main topic of said sentence;

removing words from said sentence that are applicable to said first set of thematic assessments;

determining, from said thematic tags, whether a second set of thematic assessments exist, said second set of thematic assessments identify words that include contextual elements remaining in said sentence but that are not essential to conveying said main topic; and

removing words from said sentence that are applicable to said second set of thematic assessments, wherein words that exist define said main topic of said sentence.

33. A computer readable medium for storing instructions, which when executed by a computer, causes the computer to perform the steps of:

generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions that define content for said input discourse; and

extracting, for each sentence in said input discourse, words that identify at least one of a plurality of major thematic points based on said thematic tags, wherein said words extracted identify content for each sentence in said input discourse.

34. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise thematic head words.

35. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise proper names.

36. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise descriptions.

37. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise named topics.

38. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise topic indicators.

39. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise definitions.

40. The computer readable medium as set forth in claim 33, wherein at least one of a plurality of major thematic points comprise developed proper names.

41. An apparatus for determining themes in an input discourse, said apparatus comprising:

means for specifying a plurality of thematic constructions that define content of discourse;

means for testing each of said thematic constructions with said input discourse to identify applicability of each of said thematic constructions to said input discourse; and

means for generating a thematic context output for said input discourse by generating, for each word in said input discourse, a thematic tag to indicate applicability of each of said thematic constructions to said input discourse.

42. An apparatus for determining thematic context of an input discourse, said apparatus comprising:

means for generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions to thematic units of said input discourse; and

means for searching said thematic context output for pre-defined combinations of said thematic tags; and

means for extracting, in combination, said thematic units corresponding to said predefined combination of said thematic tags to extract a theme for said input stream.

43. An apparatus for generating kernel sentences from input discourse, said apparatus comprising:

means for generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions including a plurality of thematic contextual elements for said input discourse;

means for accessing said thematic tags in said thematic context output; and

means for removing words from each sentence of said input discourse if a thematic tag for a corresponding word indicates applicability to one of said thematic contextual elements, wherein words that exist define a kernel sentence.

44. An apparatus for determining a topic for each sentence in an input discourse, said apparatus comprising:

means for generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions that define content for said input discourse; and

means for accessing said thematic tags in said thematic context output; and

means for extracting a main topic for said sentence based on said thematic tags including extracting context words that support said main topic.

45. An apparatus for extracting content in an input discourse, said apparatus comprising:

means for generating a thematic context output for said input discourse comprising a plurality of thematic tags, wherein said thematic tags indicate applicability of a plurality of thematic constructions that define content for said input discourse; and

means for extracting, for each sentence in said input discourse, words that identify at least one of a plurality of major thematic points based on said thematic tags, wherein said words extracted identify content for each sentence in said input discourse.