Systems and methods for user-interest sensitive condensation

• US 7,801,723 B2
• Filed: 11/30/2004
• Issued: 09/21/2010
• Est. Priority Date: 11/30/2004
• Status: Expired due to Fees

• Current Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Original Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Inventors: Kaplan, Ronald M., Bobrow, Daniel G., Crouch, Richard
• Primary Examiner(s): Hudspeth; David R
• Assistant Examiner(s): Spooner; Lamont M
• Application Number:
  US10/999,792
• Publication Number:
  US 20060116860A1
• Time in Patent Office: 2,121 Days
• Field of Search: 704/9, 704/1, 704/10, 707 2- 6, 707/736, 707/767, 715/254
• US Class Current: 704/9
• CPC Class Codes: G06F 40/258 Heading extraction; Automatic titling; Numbering
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Techniques are presented to determine user-interest sensitive condensations of a passage. One or more passages are selected and user interest information, condensation transformations and optional meaning distortion constraints are identified. The foci of user interest within the selected passages are determined based the similarity of the elements in the selected passages to elements in the user interest information. The condensation transformations are applied to the selected passages to preferentially retain user foci while eliding less salient information. The resultant condensate provides signals the user-interest sensitive meaning of the passage. Meaning distortions constraints are optionally applied in conjunction with the condensation transformations or in creating the condensation transformations to reduce the likelihood of distorting the meaning of the passage.

Claims

1. A method of user-interest sensitive text condensation comprising the steps of:

selecting a text passage comprised of elements by highlighting text or by inputting text;

determining a language characteristic using XML or HTML, linguistic analysis of the text passage, or a method determining language;

determining user interest information using at least one of entered keywords, monitored Internet browsing patterns, user academic information and user professional information;

determining a parsing grammar for the determined language based on at least one of determined language characteristic, genre of the selected text, and characteristic of the selected passage;

determining a generation grammar;

determining a meaning structure based on the selected text passage;

determining condensation transformations comprised of rewrite rules which delete, merge and change elements of the meaning structure;

determining meaning distortion constraints beyond which the meaning structure is identified as being distorted after the condensation transformations are applied;

determining a packed meaning structure based on the selected passage and the determined parsing grammar;

determining a user foci of interest from within the packed meaning structure based on the user interest information;

determining a reduced meaning structure based on the condensation transformations, the packed structure, the user foci, and the meaning distortion constraints;

determining at least one candidate condensation structure based on the reduced meaning structure and a predictive model or a statistical disambiguation model;

determining a text condensation based on candidate condensation structures and the determined generation grammar; and

selecting a user-interest sensitive text condensation from the candidate text condensates.

2. The method of claim 1, in which determining the condensation further comprises the step of prioritizing the retention of the elements of the passage that match elements of the user interest information.

3. The method of claim 2, in which the matches are determined based on at least one of:

literal, distributional, statistical, and co-occurrence similarity.

4. The method of claim 3, in which the conceptual similarity is based on at least one of:

a lexicon;

an ontology;

statistics and a machine-readable information source.

5. The method of claim 1, in which the meaning structure is determined based on a parsing grammar.

6. The method of claim 5, in which the generation grammar is determined based on at least one of:

a characteristic of the passage and user preference.

7. The method of claim 6, in which the characteristic of the passage is at least one of:

a language and a genre characteristic.

8. The method of claim 1, in which the packed meaning structure is comprised of grammatical sentences.

9. The method of claim 1, in which the passage is at least one sentence.

10. The method of claim 9, in which the at least one sentence is comprised of at least two words.

11. The method of claim 1, in which the passage is converted into meaning structures using a parsing grammar.

12. The method of claim 1, in which the meaning distortion constraints are useable to adjust the condensation of meaning structures characterized as downward monotonic.

13. The method of claim 12, in which the meaning distortion constraint adjustment is at least one of:

preventing condensation and repairing meaning in the meaning structure.

14. The method of claim 1, in which the determined condensate signals the meaning of the passage.

15. The method of claim 1, in which determining the condensation further comprises the steps of:

resolving non-local references in the passage to augment local elements with the non-local co-reference information for the purpose of prioritizing the retention of the elements of the passage that match elements of the user interest information.

16. A system for user-interest sensitive text condensation comprising:

a processor that selects a text passage by highlighting the text or by inputting the text;

user interest information received from at least one of entered keywords, monitored Internet browsing patterns, and user academic or professional information;

at least one condensation transformation expressed as rewrite rules and functions which reduces the selected passage by merging, deleting and changing the elements of the selected passage to remove modifiers and less salient information;

a meaning structure determined by using a parsing grammar;

a reduced meaning structure derived from the selected passage and based on input user information;

meaning distortion constraints beyond which the meaning structure is identified as being distorted after the at least one condensation transformation is applied;

a plurality of candidate condensates based on the reduced meaning structure which are optimally ranked based on a predictive model or a statistical disambiguation model; and

a user-interest sensitive condensation selected from the candidate condensates based on the input user interests.

17. The system of claim 16, in which elements of the passage that match elements of the user interest information are preferentially retained.

18. The system of claim 17, in which the matches are determined based on at least one of:

literal, conceptual, distributional, statistical, and co-occurrence similarity.

19. The system of claim 18, in which the conceptual similarity is based on a lexicon;

an ontology;

statistics and a machine-readable information source.

20. The system of claim 16, in which the processor determines a meaning structure for the passage;

and further determines the condensate based on elements of the meaning structure that match elements of the user interest information.

21. The system of claim 20, in which the meaning structure is a packed meaning structure.

22. The system of claim 21, in which the packed meaning structure is at least one of:

a logical formula in the predicate calculus, a minimal recursion semantic structure, a head-driven phrase structure and an f-structure.

23. The system of claim 16, in which the parsing grammar is determined based on at least one of:

a characteristic of the passage and user preference.

24. The system of claim 23, in which the characteristic of the passage is at least one of:

a language and a genre characteristic.

25. The system of claim 16, in which the condensate is comprised of grammatical sentences.

26. The system of claim 16, in which the passage is at least one sentence.

27. The system of claim 26, in which the at least one sentence is comprised of at least two words.

28. The system of claim 16, in which the passage is converted into at least one meaning structure using a parsing grammar.

29. The system of claim 16, in which the condensation rules are re-write rules.

30. The system of claim 16, in which the passage is converted into a packed meaning structure and the condensation transformations are applied to the packed structure based on meaning distortion constraints.

31. The system of claim 30, in which the meaning distortion constraints adjust the condensation of the meaning structures characterized as downward monotonic.

32. The system of claim 31, in which the meaning distortion constraint adjustment is at least one of:

preventing condensation and adjusting the meaning in the meaning structure.

33. A method of creating a user-interest sensitive text condensation comprising the steps of:

determining a passage by highlighting text or by entering text;

determining user interest information by at least one of entered keywords, monitored Internet browsing patterns, and user academic or professional information;

determining condensation transformations expressed as rewrite rules and functions;

determining at least one reduced passage based on user interest information;

determining meaning distortion constraints beyond which the at least one reduced passage is identified as being distorted after the condensation transformations are applied;

determining at least one text condensate based on comparison between the reduced passage, the user interest information, and the condensation transformations; and

selecting one user-interest sensitive text condensation from the at least one reduced passage.

34. A method of using a computer to generate a user-interest sensitive text condensation, the method comprising:

selecting a text input passage by use of a computer to highlight text or input text;

parsing the input passage into elements using a computer processor;

determining user foci based on similarities between the user information and the elements of the selected text input passage;

determining a condensation transformation expressed as rewrite rules and functions which reduce the selected passage by merging, deleting, and changing the elements of the selected passage;

determining a reduced passage text condensation by removing information from the passage based on the condensation transformation, the user foci, and the user interest information; and

determining meaning distortion constraints beyond which the reduced passage text condensation is identified as being distorted after the condensation transformation is applied to the passage.

35. A computer readable storage medium comprising computer readable program code embodied on the computer readable storage medium, the computer readable program code usable to program a computer for user-interest sensitive condensation comprising the steps of:

determining a passage by highlighting one or more multi-word sentences through use of a computer mouse controlled cursor wherein a mouse button is held down as the cursor is dragged over a selected passage in order to capture the passage;

determining user interest information by parsing the text of the user selected passage to determine alternative meaning structures associated with the passage;

determining condensation transformations expressed as rewrite rules and functions which reduce the selected passage by deleting, merging, and/or adjusting the element of alternative meaning structures in order to compare user interest information with alternative meaning structures;

determining meaning distortion constraints beyond which the alternative meaning structures are identified as being distorted after the condensation transformations are applied; and

applying a condensation transformation based on the determined passage, and the user interest information to condense facts in the alternative meaning structures to create condensed packed meaning structure.

36. A user-interest sensitive condensation derived by using a computer processor to transform a user selected passage of a text into a packed meaning structure which is at least one of:

a logical formula in the predicate calculus,

a minimal recursion semantic structure,

a head-driven phrase structure, and

an f-structure; and

wherein the user note and

is based on one or more input indicators of user interest and includes;

a condensation transformation expressed as rewrite rules and functions which reduce the selected passage by merging, deleting, and changing elements of the selected passage;

a reduced passage text note by removing information from the passage based on the condensation transformation, the user interest, and user interest information; and

meaning distortion constraints beyond which the reduced passage text condensation is identified as being distorted after the condensation transformation is applied to the passage,

wherein the meaning structure for a passage is created through the use of a stochastic disambiguation model and/or predictive model which assigns higher probabilities to better examples from a training set and assigns lower probabilities to less desirable or less appropriate examples.

37. The user-interest sensitive condensation of claim 36, further comprising at least one selectable user interface component operable to adjust the display of a portion of the user-interest sensitive condensation.

38. The user-interest sensitive condensation of claim 37, wherein the displayed portion of the user-interest sensitive condensation is associated with the passage.

39. The user-interest sensitive condensation of claim 37, wherein the user selectable interface component is at least one of:

a button, a link and a uniform resource locator.

40. The user-interest sensitive condensation of claim 36, wherein the condensate contains information from the passage.

41. The user-interest sensitive condensation of claim 36, wherein the passage is displayed in at least one of:

a window;

a dialog box;

a pop-up-box and a status area.

42. The user-interest sensitive condensation of claim 36, wherein the selectable user interface component is responsive to at least one of:

voice, touch and eye-gaze input.

43. The user-interest sensitive condensation of claim 36, wherein the passage is a selected portion of a word processor.

44. The user-interest sensitive condensation of claim 43, in which the condensation is displayed contextually adjacent with the passage.

45. The user-interest sensitive condensation of claim 44, in which the contextually adjacent location is at least one of:

a document margin, a fluid text display, and a window.

46. A method of creating a corrected user-interest sensitive condensation comprising the steps of:

determining a user selected text passage by highlighting or inputting one or more multi-word sentences;

parsing the user selected text into elements using a parsing grammar;

determining user interest information from at least one of entered keywords, monitored Internet browsing patterns, user academic information and user professional information;

determining condensation transformations expressed as rewrite rules and functions;

determining non-local references in the passage;

determining a text condensation based on comparison between the determined passage, the user interest information, and the condensation transformations;

determining meaning distortion constraints beyond which the text condensation is identified as being distorted after the application of at least one condensation transformation;

identifying dropped words and meaning distortions within the text condensation; and

producing a corrected text condensation based on the text condensation and the addition of dropped words and repairing of meaning distortions.