Assessing Trust in Information Sources

• US 20120066167A1
• Filed: 09/13/2010
• Published: 03/15/2012
• Est. Priority Date: 09/13/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Srivatsa, Mudhakar, Fokoue, Achille
• Granted Patent: US 8,825,577 B2
• Time in Patent Office: Days
• US Class Current: 706/52
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation

Abstract

Techniques for assessing trust in information are provided. The techniques include collecting information from one or more information sources, detecting one or more inconsistencies in one or more statements from the information collected from the one or more sources, assigning a penalty to each information source involved in an inconsistency and a reward for each statement not involved in any inconsistencies, and using each assigned penalty and reward to perform an aggregation to assess trust in the collected information.

Claims

1. A method for assessing trust in information, wherein the method comprises:

collecting information from one or more information sources;

detecting one or more inconsistencies in one or more statements from the information collected from the one or more sources;

assigning a penalty to each information source involved in an inconsistency and a reward for each statement not involved in any inconsistencies; and

using each assigned penalty and reward to perform an aggregation to assess trust in the collected information.

2. The method of claim 1, wherein the one or more information sources are heterogeneous with respect to trustworthiness.

3. The method of claim 1, wherein the information is represented as at least one of a simple first order logic statement, and a statement from a probabilistic extension of first order logic in order to capture uncertain information.

4. The method of claim 1, wherein an inconsistency is represented as a justification, wherein a justification comprises a minimal set of one or more axioms responsible for an inconsistency.

5. The method of claim 4, wherein a justification is weighted based on its likelihood of being selected if a selection process was random.

6. The method of claim 1, wherein using each assigned penalty and reward to perform an aggregation to assess trust in the collected information comprises providing each penalty and reward to a trust model for performance of an aggregation.

7. The method of claim 1, wherein a trust assessment of an information source is based on whether one or more statements of the source are among the one or more detected inconsistencies.

8. The method of claim 1, further comprising storing information in a knowledge base that captures complexity of the information through an encoding to a decidable subset of first order logic, and uncertainty of the information.

9. The method of claim 8, using trust assessment to refine the knowledge base.

10. The method of claim 1, further comprising using trust assessment to enhance information quality via determining a strength of credibility for each information source.

11. The method of claim 1, further comprising using trust assessment to classify one or more information sources wherein an inconsistency involves two or more information sources.

12. The method of claim 1, further comprising using trust assessment to classify one or more information sources wherein a penalty is asymmetrically assigned to an information source based on prior trust of the information source.

13. The method of claim 12, wherein a less trustworthy source is assigned a higher penalty.

14. The method of claim 1, wherein detecting one or more inconsistencies in one or more statements from the information collected from the one or more sources comprises the steps of:

representing a problem of justification computation as a traversal of a search tree; and

uniformly and randomly selecting one or more justifications from the search tree without performing a full traversal.

15. The method of claim 14, wherein uniformly and randomly selecting one or more justifications from the search tree without performing a full traversal does not produce a uniform and random sample of justifications, and a tradeoff between computation cost and bias is controlled by a tunable parameter.

16. The method of claim 1, further comprising providing a system, wherein the system comprises one or more distinct software modules, each of the one or more distinct software modules being embodied on a tangible computer-readable recordable storage medium, and wherein the one or more distinct software modules comprise a knowledgebase module, an inconsistency detection module, a weighted justification generation module and a trust computation model module executing on a hardware processor.

17. A computer program product comprising a tangible computer readable recordable storage medium including computer useable program code for assessing trust in information, the computer program product including:

computer useable program code for collecting information from one or more information sources;

computer useable program code for detecting one or more inconsistencies in one or more statements from the information collected from the one or more sources;

computer useable program code for assigning a penalty to each information source involved in an inconsistency and a reward for each statement not involved in any inconsistencies; and

computer useable program code for using each assigned penalty and reward to perform an aggregation to assess trust in the collected information.

18. The computer program product of claim 17, wherein an inconsistency is represented as a justification, wherein a justification comprises a minimal set of one or more axioms responsible for an inconsistency.

19. The computer program product of claim 17, further comprising computer useable program code for using trust assessment to classify one or more information sources wherein a penalty is asymmetrically assigned to an information source based on prior trust of the information source.

20. The computer program product of claim 17, wherein the computer useable program code for detecting one or more inconsistencies in one or more statements from the information collected from the one or more sources comprises:

computer useable program code for representing a problem of justification computation as a traversal of a search tree; and

computer useable program code for uniformly and randomly selecting one or more justifications from the search tree without performing a full traversal.

21. A system for assessing trust in information, comprising:

a memory; and

at least one processor coupled to the memory and operative to;

collect information from one or more information sources;

detect one or more inconsistencies in one or more statements from the information collected from the one or more sources;

assign a penalty to each information source involved in an inconsistency and a reward for each statement not involved in any inconsistencies; and

use each assigned penalty and reward to perform an aggregation to assess trust in the collected information.

22. The system of claim 21, wherein an inconsistency is represented as a justification, wherein a justification comprises a minimal set of one or more axioms responsible for an inconsistency.

23. The system of claim 21, wherein the at least one processor coupled to the memory is further operative to use trust assessment to classify one or more information sources wherein a penalty is asymmetrically assigned to an information source based on prior trust of the information source.

24. The system of claim 21, wherein the at least one processor coupled to the memory operative to detect one or more inconsistencies in one or more statements from the information collected from the one or more sources is further operative to:

represent a problem of justification computation as a traversal of a search tree; and

uniformly and randomly select one or more justifications from the search tree without performing a full traversal.

25. An apparatus for assessing trust in information, the apparatus comprising:

means for collecting information from one or more information sources;

means for detecting one or more inconsistencies in one or more statements from the information collected from the one or more sources;

means for assigning a penalty to each information source involved in an inconsistency and a reward for each statement not involved in any inconsistencies; and

means for using each assigned penalty and reward to perform an aggregation to assess trust in the collected information.