System and method for evidence accumulation and hypothesis generation

• US 7,421,419 B2
• Filed: 04/12/2006
• Issued: 09/02/2008
• Est. Priority Date: 04/12/2005
• Status: Expired due to Fees

• Current Assignee: Viziant Corp.
• Original Assignee: Viziant Corp.
• Inventors: Maren, Alianna J.
• Primary Examiner(s): Holmes; Michael B
• Application Number:
  US11/279,465
• Publication Number:
  US 20070005523A1
• Time in Patent Office: 874 Days
• Field of Search: None
• US Class Current: 706/52
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 5/046 Forward inferencing; Production systems

Abstract

A methodology, a system, and an apparatus is defined for performing evidence-based decision-making about matching a given entity against one or more of a set of known or reference entities. A satisfactory decision is achieved as a function of both potentiality and plausibility, where plausibility refers to the full set of values garnered by the evidence accumulation process in the process of generating belief/disbelief/uncertainty/conflict masses. Potentiality is a mechanism to set the various match threshold values, where the thresholds define acceptable confidence levels for decision-making. Evidence is computed on the basis of partial matching of feature vector elements, where separate and distinct feature vectors are associated with both the given entity and each of the reference entities. Following evidence-combination methods, evidence is accrued for both the positive and negative decisions regarding a potential match.

Claims

1. A system for performing evidence-based decision-making comprising:

an evidence processor, configured to match an extracted entity against a set of reference entities;

an evidence selection processor for defining a rule set to be applied to the extracted entity and the plurality of reference entities and configured to generate a total evidence function;

a threshold processor configured to generate a context-dependent threshold for an acceptable decision, wherein the acceptable decision is achieved when a value of the total evidence function exceeds a predetermined threshold value; and

a decision processor, configured to compare the total evidence function to the context-dependent threshold and determine whether to accumulate and aggregate further evidence or to generate a decision-result; and

wherein the system provides a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

2. A system for performing evidence-based decision-making as claimed in claim 1, wherein the evidence processor is configured to further accumulate and aggregate a set of enhanced reference entities against which each of the extracted entity will be matched;

and further comprising one or more data sources, configured to enhancements in the form of either feature vector elements and/or corrections to feature vector elements for either or both the reference entities and the extracted entities, thereby generating a plurality of augmented feature vectors for each of the plurality of extracted entities and/or the reference entities.

3. A system for performing evidence-based decision-making as claimed in claim 1, wherein the evidence selection processor further comprises:

a hypothesis generator configured to generate one or more hypotheses about the extracted entity; and

a hypothesis validation processor configured to validate, refute and/or eliminate the generated hypotheses to generate a total evidence function which is computed uniquely and distinctively for each hypothesis regarding a potential match between an extracted entity and a reference entity.

4. A system for performing evidence-based decision-making as claimed in claim 3, wherein the rule set is used both by the hypothesis generator to generate a plurality of hypotheses and by the hypothesis validation processor to gather data to validate or refute the generated hypotheses.

5. A system for performing evidence-based decision-making as claimed in claim 1, wherein the threshold processor is further configured to:

use the rule set to define the context-dependent threshold for an acceptable decision using a potentiality calculation and plausibility calculation.

6. A system for performing evidence-based decision-making, as claimed in claim 1, further comprising:

a reconciliation processor, configured to;

acquire additional evidence from the one or more data sources, according to the rule set provided by the evidence selection processor, for the extracted entity and each of its associated hypothetical matches to a reference entity or enhanced reference entity;

evaluate the hypothetical matches and determine if additional evidence is required to evaluate the hypothetical matches; and

a second evidence processor by which the additional evidence is aggregated with the existing evidence associated with the extracted entity, forming an expanded feature vector set uniquely associated with the extracted entity.

7. A system for performing evidence-based decision-making as claimed in claim 1, wherein the evidence processor is configured to accumulate evidence so that the evidence is traceable.

8. A system for performing evidence-based decision-making as claimed in claim 1, wherein each of the extracted entities and reference entities further comprise a plurality of feature vectors.

9. A system for performing evidence-based decision-making as claimed in claim 8, wherein at least one of the feature vectors of the reference entities is not fully populated with a plurality of element values.

10. A system for performing evidence-based decision-making as claimed in claim 8, wherein some and/or each of the plurality of feature vectors comprises a plurality of vector elements.

11. A system for performing evidence-based decision-making as claimed in claim 8, wherein the evidence processor is configured to obtain additional element values for the plurality of reference entity feature vectors.

13. A system for performing evidence-based decision-making from a set of data elements as claimed in claim 1, further comprising an evidence selection processor for defining a plurality of rule sets to be applied to each of the plurality of extracted entities and the plurality of reference entities and configured to generate the initial set of extracted entities.

14. A system for performing evidence-based decision-making from a set of data elements as claimed in claim 13, wherein the evidence selection processor further comprises:

a hypothesis generator configured to generate one or more hypotheses about each of the plurality of extracted entities; and

a hypothesis validation processor configured to validate, refute and/or eliminate the generated hypotheses to generate the initial set of extracted entities.

15. A system for performing evidence-based decision-making as claimed in claim 14, wherein the rule set is used both by the hypotheses generator to generate a plurality of hypotheses and by the evidence processor to gather data to validate or refute the generated hypotheses.

18. A system for performing evidence-based decision-making from a set of data elements, as claimed in claim 13, wherein the rule set is derived by performing a multidimensional lookup task and characterizing the plurality of vector elements of the plurality of feature vectors against normative values.

19. A system for performing evidence-based decision-making from a set of data elements, as claimed in claim 13, further comprising:

a reconciliation processor, configured to acquire additional evidence from the one or more data sources, according to the rule set provided by the evidence selection processor, for each member of the candidate positive pool and each of its associated hypothetical matches to a reference entity or enhanced reference entity;

a second evidence processor by which the additional evidence is aggregated with the existing evidence associated with a candidate positive entity, forming an expanded feature vector set uniquely associated with that candidate positive entity, and

a decision processor, configure to apply an evaluation to the new feature vector associated with a given candidate positive entity, to further refine the previous decision as to whether the candidate positive is a definite match, a definite not-match, or requires more evidence for hypothesis refutation regarding its hypothesized match against a enhanced reference entity or a reference entity.

12. A system for performing evidence-based decision-making from a set of data elements comprising:

a plurality of extracted entities about which a decision will be made for each extracted entity, wherein each extracted entity has associated with it a plurality of feature vectors each having a plurality of feature vector elements;

a set of reference entities against which each of the plurality of extracted entities will be matched;

a set of enhanced reference entities against which each of the plurality of extracted entities will be matched;

one or more data sources, configured to enhancements in the form of either feature vector elements and/or corrections to feature vector elements for either or both the reference entities and the extracted entities, thereby generating a plurality of augmented feature vectors for each of the plurality of extracted entities and/or the reference entities;

an evidence processor, configured to compare the plurality of augmented feature vectors associated with each of the plurality of extracted entities against a plurality of feature vectors for the set of reference entities or the set of enhanced reference entities; and

a threshold processor configured to generate a context-dependent threshold for an acceptable decision,

wherein the system provides a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

16. A system for performing evidence-based decision-making as claimed in claim 12, wherein the threshold processor is further configured to:

use the rule set to define the context-dependent threshold for an acceptable decision using a potentiality calculation and plausibility calculation.

17. A system for performing evidence-based decision-making from a set of data elements, as claimed in claim 12, wherein the set of enhanced reference entities include various permutations of the plurality of reference entities, and wherein the set of enhanced reference entities is larger than and inclusive of the set of reference entities.

20. A system for performing evidence-based decision-making comprising:

an evidence processor, configured to compare a plurality of augmented feature vectors associated with each of a plurality of extracted entities against a plurality of feature vectors for a set of reference entities or a set of enhanced reference entities; and

a threshold processor configured to generate a context-dependent threshold for an acceptable decision, wherein the acceptable decision is achieved when a value of a total evidence function exceeds a predetermined threshold value,

wherein the system provides a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

21. A system for performing evidence-based decision-making as claimed in claim 20, further comprising one or more data sources, configured to provide data for the set of reference entities and the set of enhanced reference entities in order to augment the plurality of feature vectors associated with each of the plurality of extracted entities, thereby generating a plurality of augmented feature vectors for each of the plurality of extracted entities.

22. A system for performing evidence-based decision-making as claimed in claim 20, further comprising:

a reconciliation processor, configured to acquire additional evidence from the one or more data sources, according to the rule set provided by the evidence selection processor, for each member of the initial set of extracted entities and each of its associated hypothetical matches to a reference entity or enhanced reference entity;

a second evidence processor by which the additional evidence is aggregated with the existing evidence associated with the extracted entity, forming an expanded feature vector set uniquely associated with that extracted entity, and

a decision processor, configure to apply an evaluation to the new feature vector associated with the extracted entity, to further refine the previous decision as to whether the candidate positive is a definite match, a definite not-match, or requires more evidence for hypothesis refutation regarding its hypothesized match against a enhanced reference entity or a reference entity.

23. A system for performing evidence-based decision-making from a set of data elements as claimed in claim 20, further comprising an evidence selection processor for defining a rule set to be applied to the extracted entity and the plurality of reference entities and configured to generate an initial set of extracted entities.

24. A system for performing evidence-based decision-making from a set of data elements as claimed in claim 23, wherein the evidence selection processor further comprises:

a hypothesis generator configured generate one or more hypotheses about the extracted entity; and

a hypothesis validation processor configured to validate, refute and/or eliminate the generated hypotheses to generate an initial set of extracted entities.

25. A system for performing evidence-based decision-making as claimed in claim 23, wherein the rule set is used both by the hypotheses generator to generate a plurality of hypotheses and by the evidence processor to gather data to validate or refute the generated hypotheses.

26. A system for performing evidence-based decision-making from a set of data elements, as claimed in claim 23, wherein the rule set is derived by performing a multidimensional lookup task and characterizing the plurality of vector elements of the plurality of feature vectors against normative values.

27. A system for performing automated security screening using evidence-based decision-making comprising:

an evidence processor, configured to match an extracted entity against a set of reference entities;

an evidence selection processor for defining a rule set to be applied to the extracted entity and the plurality of reference entities and configured to generate a total evidence function;

a threshold processor configured to generate a context-dependent threshold for an acceptable decision; and

a decision processor, configured to compare the total evidence function to the context-dependent threshold and determine whether to accumulate and aggregate further evidence or to generate a decision,

wherein the system provides a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

28. A system for performing evidence-based decision-making as claimed in claim 27, wherein for each extracted entity there exists a primary information vector, an activity vector and a context vector.

29. A system for performing evidence-based decision-making as claimed in claim 27, wherein the primary information vector consists of personal information, including but not limited to a name and a date of birth of the extracted entity.

30. A system for performing evidence-based decision-making as claimed in claim 27, wherein the activity vector may include flight information, public appearance information and/or statement information.

31. A system for performing evidence-based decision-making as claimed in claim 27, wherein the context vector includes but is not limited to contextual information related to the information contained in the activity vector.

32. A system for performing evidence-based decision-making comprising a processor configured to:

match an extracted entity against a set of reference entities;

define a rule set to be applied to the extracted entity and the plurality of reference entities and configured to generate a total evidence function;

generate a context-dependent threshold for an acceptable decision, wherein the context-dependent threshold is a function of a potentiality and plausibility value; and

compare the total evidence function to the context-dependent threshold and determine whether to accumulate and aggregate further evidence or to generate a decision-result,

wherein the system provides a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

33. A system for performing evidence-based decision-making as claimed in claim 32, wherein the potentiality value is independent of any information about the extracted entity and is a measure representative of a need to obtain more information about the extracted entity.

34. A system for performing evidence-based decision-making as claimed in claim 32, wherein the plausibility value is a degree to which it is credible that a given assertion could possibly be true.

35. A method for performing evidence-based decision-making comprising the steps of:

matching an extracted entity against a set of reference entities;

defining a rule set to be applied to the extracted entity and the plurality of reference entities and configured to generate a total evidence function;

generating a context-dependent threshold for an acceptable decision, wherein the acceptable decision is achieved when a value of the total evidence function exceeds a predetermined threshold value;

comparing the total evidence function to the context-dependent threshold and determine whether to accumulate and aggregate further evidence or to generate a decision-result; and

providing a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

36. A method for performing evidence-based decision-making from a set of data elements comprising the steps of:

identifying a plurality of extracted entities about which a decision will be made for each extracted entity, wherein each extracted entity has associated with it a plurality of feature vectors each having a plurality of feature vector elements;

identifying a set of reference entities against which each of the plurality of extracted entities will be matched;

identifying a set of enhanced reference entities against which each of the plurality of extracted entities will be matched;

generating a plurality of augmented feature vectors for each of the plurality of extracted entities and/or the reference entities from one or more data sources, configured to enhancements in the form of either feature vector elements and/or corrections to feature vector elements for either or both the reference entities and the extracted entities, thereby;

comparing the plurality of augmented feature vectors associated with each of the plurality of extracted entities against a plurality of feature vectors for the set of reference entities or the set of enhanced reference entities; and

generating a context-dependent threshold for an acceptable decision,

providing a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

37. A method for performing evidence-based decision-making comprising the steps of:

comparing a plurality of augmented feature vectors associated with each of a plurality of extracted entities against a plurality of feature vectors for a set of reference entities or a set of enhanced reference entities; and

generating a context-dependent threshold for an acceptable decision, wherein the acceptable decision is achieved when a value of a total evidence function exceeds a predetermined threshold value,

providing a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.

38. A method for performing automated security screening using evidence-based decision-making comprising the steps of:

matching an extracted entity against a set of reference entities;

defining a rule set to be applied to the extracted entity and the plurality of reference entities and configured to generate a total evidence function;

generating a context-dependent threshold for an acceptable decision; and

comparing the total evidence function to the context-dependent threshold and determine whether to accumulate and aggregate further evidence or to generate a decision,

providing a degree of validation or refutation of the extracted entity matching one or more of the set of reference entities.