Scoring concepts for contextual personalized information retrieval

• US 8,463,810 B1
• Filed: 06/01/2007
• Issued: 06/11/2013
• Est. Priority Date: 06/01/2006
• Status: Active Grant

• Current Assignee: Monster Worldwide Incorporated (Randstad NV)
• Original Assignee: Monster Worldwide Incorporated (Randstad NV)
• Inventors: Rennison, Earl
• Primary Examiner(s): Ehichioya, Fred I
• Assistant Examiner(s): OWYANG, MICHELLE N
• Application Number:
  US11/757,040
• Time in Patent Office: 2,202 Days
• Field of Search: None
• US Class Current: 707/771
• CPC Class Codes: G06F 16/319 Inverted lists
• CPC Class Codes: G06F 16/3332 Query translation
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/3335 Syntactic pre-processing, e.g. stopword elimination, stemming
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06F 16/906 Clustering; Classification
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Information retrieval systems face challenging problems with delivering highly relevant and highly inclusive search results in response to a user'"'"'s query. Contextual personalized information retrieval uses a set of integrated methodologies that can combine automatic concept extraction/matching from text, a powerful fuzzy search engine, and a collaborative user preference learning engine to provide accurate and personalized search results. The system can include constructing a search query to execute a search of a database. The system can parse an input query from a user conducting the search of the database into sub-strings, and can match the sub-strings to concepts in a semantic concept network of a knowledge base. The system can further map the matched concepts to criteria and criteria values that specify a set of constraints on and scoring parameters for the matched concepts.

Claims

1. A computer-implemented method for generating search results from an input query, the method comprising:

using a computer system to execute method steps comprising;

receiving an input query for a search comprising target concepts;

determining a structured set of criteria associated with criteria values, wherein the input query is normalized into the structured set of criteria, and wherein the criteria specify how to filter and score the target concepts, and wherein the criteria value is a tuple that corresponds to the criteria;

mapping the structured set of criteria and associated criteria values to nodes in a knowledge base comprising a semantic network relationship among concepts and an index of a plurality of documents, wherein each node represents a target concept associated with a target value, and wherein the target concept is an instance of a category;

matching nodes for the target values of the target concepts to nodes for selection aspects of the criteria and criteria values using transitivity across the semantic network between these nodes;

selecting the set of target concepts having the associated target values that match the selection aspects of the criteria and the criteria values in the matching of the nodes;

computing a partial score on each of the target values in the selected target concepts based on scoring aspects of the criteria and criteria values, each partial score scoring each of the target values against the criteria values as descendants of the criteria values in the semantic network, and each partial score measuring a degree of match between the target values used in computing that partial score and the criteria values on which that partial score is based;

computing at least one weight for each of the criteria values;

combining the partial scores for the target values that matched a given criteria value to generate a single partial score for that criteria value that is based on the at least one weight for that criteria value, wherein each selected target concept has a set of single partial scores for each criteria value;

determining a total score for each selected target concept by integrating the set of single partial scores of the selected target concept; and

applying the total scores for the target concepts to generate search results for the user in a ranked order, the search results including one or more of the documents indexed.

2. The method of claim 1, wherein the input query is a query for job descriptions.

3. The method of claim 1, wherein the input query is a query for ré

sumé

s for job candidates.

4. The method of claim 1, wherein selecting the set of target concepts further comprises using a constraint tree to construct a query to select the set, the constraint tree including a hierarchy of constraints.

5. The method of claim 1, wherein the criteria are selected from a group consisting of:

filter criteria specifying filtering of search results based on the presence of the matching concepts, include criteria specifying concepts that are required to be present in search results, score criteria specifying how individual concepts in search results are scored, search criteria specifying how sub-sets of concepts in search results are selected and scored as a group, and any combination of these.

6. The method of claim 5, wherein the filter criteria are instantiated as filter criteria values, the include criteria are instantiated as include criteria values, the search criteria are instantiated as search criteria values, and the score criteria are instantiated as score criteria values.

7. The method of claim 6, wherein selecting a set of target concepts further comprises:

selecting a subset of target concepts for each filter criteria value using filter criteria value parameters; and

computing inclusion and exclusion filter sets by combining the subsets together using set operations specified by the filter criteria values.

8. The method of claim 6, wherein selecting a set of target concepts further comprises:

selecting a subset of target concepts for each include criteria value using include criteria value parameters; and

adding the selected target concepts to a search result set, applying inclusion and exclusion filter sets when specified.

9. The method of claim 6, wherein selecting a set of target concepts further comprises:

selecting a subset of target concepts for each search criteria value using search criteria value parameters;

computing a partial score for each of the target concepts in the subset; and

adding each of the selected target concepts to a search result set.

10. The method of claim 6, further comprising computing, for each of the target values in the selected target concepts, a partial score based on the score criteria value for the selected target concepts in a search result set, wherein the search result set was created using the include criteria values and the search criteria values.

11. The method of claim 1, further comprising using an evaluator that is a function that takes as input a target value and produces an output in a fixed range.

12. The method of claim 11, wherein the evaluator is a curve function.

13. The method of claim 12, wherein the curve function is defined by a set of piece wise contiguous Bezier Curves.

14. The method of claim 1, further comprising using an evaluator that is a function that takes as input values defined by one or more attributes or attribute paths from a target category and computes a partial score based on those values.

15. The method of claim 14, wherein the evaluator computes a similarity measure by computing a comparison of two basis vectors that have components that are determined by a transitive closure constrained by an attribute path.

16. The method of claim 14, wherein the evaluator computes a similarity measure by computing a dot product of two basis vectors that have components with lengths based on a frequency of concepts in an attribute path.

17. The method of claim 14, wherein the evaluator computes a similarity measure by computing a dot product of two basis vectors that have components, the components determined by values selected along an attribute path that has transitivity through a primary constraining attribute path element.

18. The method of claim 1, wherein the partial score computed on each of the target values is equal to a default score used when a target value is not available.

19. The method of claim 6, further comprising computing, for each of the score criteria values, the partial score for each of the target values in the selected target concepts.

20. The method of claim 1, wherein the weights further comprise criteria value weights computed for each of the criteria values using a weight model.

21. The method of claim 1, wherein the weight is scaled and translated based on a weight range.

22. The method of claim 1, further comprising normalizing a sum of the partial scores, for the target values, weighted by the weights.

23. The method of claim 1, further comprising computing a degree of match calculation.

24. The method of claim 1, wherein the weights further comprise weights biased by user feedback.

25. The method of claim 1, wherein the search results generated for the user are presented in ranked order sorted by score.

26. The method of claim 1, wherein there are a plurality of target values corresponding to each criteria value.

27. The method of claim 26, wherein each of the plurality of target values has a partial score against that criteria value which are combined to generate a single partial score for that criteria value.

28. The method of claim 1, wherein there are a plurality of criteria values, each having a single partial score for each of the selected target concepts.

29. A computer program product having a non-transitory computer-readable storage medium having computer program instructions recorded thereon for generating search results from an input query, the computer program instruction comprising instructions for:

receiving an input query for a search comprising target concepts;

determining a structured set of criteria associated with criteria values, wherein the input query is normalized into the structured set of criteria, and wherein the criteria specify how to filter and score the target concepts, and wherein the criteria value is a tuple that corresponds to the criteria;

mapping the structured set of criteria and associated criteria values to nodes in a knowledge base comprising a semantic network relationship among concepts and an index of a plurality of documents, wherein each node represents a target concept associated with a target value, and wherein the target concept is an instance of a category;

matching nodes for the target values of the target concepts to nodes for selection aspects of the criteria and criteria values using transitivity across the semantic network between these nodes;

selecting the set of target concepts having the associated target values that match the selection aspects of the criteria and the criteria values in the matching of the nodes;

computing a partial score on each of the target values in the selected target concepts based on scoring aspects of the criteria and criteria values, each partial score scoring each of the target values against the criteria values as descendants of the criteria values in the semantic network, and each partial score measuring a degree of match between the target values used in computing that partial score and the criteria values on which that partial score is based;

computing at least one weight for each of the criteria values;

combining the partial scores for the target values that matched a given criteria value to generate a single partial score for that criteria value that is based on the at least one weight for that criteria value, wherein each selected target concept has a set of single partial scores for each criteria value;

determining a total score for each selected target concept by integrating the set of single partial scores of the selected target concept; and

applying the total scores for the target concepts to generate search results for the user in a ranked order, the search results including one or more of the documents indexed.

30. The computer program product of claim 29, wherein the input query is selected from a group consisting of:

a query for job descriptions and a query for ré

sumé

s for job candidates.

31. The computer program product of claim 29, wherein selecting the set of target concepts further comprises using a constraint tree to construct a query to select the set, the constraint tree including a hierarchy of constraints.

32. The computer program product of claim 29, wherein the criteria are selected from a group consisting of:

filter criteria specifying filtering of search results based on the presence of the matching concepts, include criteria specifying concepts that are required to be present in search results, score criteria specifying how individual concepts in search results are scored, search criteria specifying how sub-sets of concepts in search results are selected and scored as a group, and any combination of these.

33. The computer program product of claim 29, further comprising using an evaluator that is a curve function that takes as input a target value and produces an output value in a fixed range.

34. The computer program product of claim 29, further comprising using an evaluator that is a function that takes as input values defined by one or more attributes or attribute paths from a target category and computes a partial score based on those values.

35. The computer program product of claim 34, wherein the evaluator computes a similarity measure by computing a comparison of two basis vectors that have components that are determined by a transitive closure constrained by an attribute path.

36. The computer program product of claim 29, wherein the weights further comprise criteria value weights computed for each of the criteria values using a weight model.

37. The computer program product of claim 29, wherein the weight is scaled and translated based on a weight range.

38. The computer program product of claim 29, further comprising normalizing a sum of the partial scores, for the target values, weighted by the weights.

39. The computer program product of claim 29, further comprising computing a degree of match calculation.

40. A computer system for generating search results from an input query, the system comprising:

at least one processor;

a computer-readable storage medium storing software components for execution by the at least one processor, the components comprising;

a contextual search manager for;

receiving an input query for a search comprising target concepts;

determining a structured set of criteria associated with criteria values, wherein the input query is normalized into the structured set of criteria, and wherein the criteria specify how to filter and score the target concepts and wherein the criteria value is a tuple that corresponds to the criteria;

mapping the structured set of criteria and associated criteria values to nodes in a knowledge base comprising a semantic network relationship among concepts and an index of a plurality of documents, wherein each node represents a target concept associated with a target value, and wherein the target concept is an instance of a category in the knowledge base, and wherein the target value is an instance of an attribute of the category;

a criteria evaluator for;

matching nodes for the target values of the target concepts to nodes for selection aspects of the criteria and criteria values using transitivity across the semantic network between these nodes;

selecting the set of target concepts having the associated target values that match the selection aspects of the criteria and the criteria values in the matching of the nodes;

computing a partial score on each of the target values in the selected target concepts based on scoring aspects of the criteria and criteria values, each partial score scoring each of the target values against the criteria values as descendants of the criteria values in the semantic network, and each partial score measuring a degree of match between the target values used in computing that partial score and the criteria values on which that partial score is based;

computing at least one weight for each of the criteria values;

combining the partial scores for the target values that matched a given criteria value to generate a single partial score for that criteria value that is based on the at least one weight for that criteria value, wherein each selected target concept has a set of single partial scores for each criteria value;

a contextual search evaluator for;

determining a total score for each selected target concept by integrating the set of single partial scores of the selected target concept; and

applying the total scores for the target concepts to generate search results for the user in a ranked order, the search results including one or more of the documents indexed.

41. The system of claim 40, wherein the input query is selected from a group consisting of:

a query for job descriptions and a query for ré

sumé

s for job candidates.

42. The system of claim 40, wherein the criteria evaluator is further for using a constraint tree to construct a query to select the set, the constraint tree including a hierarchy of constraints.

43. The system of claim 40, wherein the criteria are selected from a group consisting of:

filter criteria specifying filtering of search results based on the presence of the matching concepts, include criteria specifying concepts that are required to be present in search results, score criteria specifying how individual concepts in search results are scored, search criteria specifying how sub-sets of concepts in search results are selected and scored as a group, and any combination of these.

44. The system of claim 40, wherein the criteria evaluator is further for using an evaluator that is a curve function that takes as input a target value and produces an output value in a fixed range.

45. The system of claim 40, wherein the criteria evaluator is further for using an evaluator that is a function that takes as input values defined by one or more attributes or attribute paths from a target category and computes a partial score based on those values.

46. The system of claim 40, wherein the weights further comprise criteria value weights computed for each of the criteria values using a weight model.

47. The system of claim 40, wherein the criteria evaluator is further for computing a degree of match calculation.