Online data fusion

Patent metadata

Field	Value
Publication number	US-9348891-B2
Application number	US-201113311034-A
Country	US
Kind code	B2
Filing date	Dec 5, 2011
Priority date	Dec 5, 2011
Publication date	May 24, 2016
Grant date	May 24, 2016

Abstract

An online data fusion system receives a query, probes a first source for an answer to the query, returns the answer from the first source, refreshes the answer while probing an additional source, and applies fusion techniques on data associated with an answer that is retrieved from the additional source. For each retrieved answer, the online data fusion system computes the probability that the answer is correct and stops retrieving data for the answer after gaining enough confidence that data retrieved from the unprocessed sources are unlikely to change the answer. The online data fusion system returns correct answers and terminates probing additional sources in an expeditious manner without sacrificing the quality of the answers.

First claim

We claim:

1. A method comprising:
receiving, by an online data fusion system comprising a processor, answers to a query from at least two probed sources in response to probing the at least two probed sources;
computing, by the processor of the online data fusion system, a probability that each answer of the answers is correct based, at least in part, upon a copying relationship between at least two of the at least two probed sources, wherein computing the probability that each answer of the answers is correct comprises
computing, by the processor, an expected probability, a maximum probability, and a minimum probability that each answer of the answers is correct, and
refreshing, by the processor, the expected probability, the maximum probability, and the minimum probability of a first answer of the answers from a first probed source of the at least two probed sources based, at least in part, on a second answer received from a second probed source of the at least two probed sources as the second answer is received from the second probed source of the at least two probed sources;

when the online data fusion system gains enough confidence that, based upon the probability that each answer of the answers is correct, probing an additional source is unlikely to change the first answer, terminating, by processor, probing without probing the additional source; and
providing, by the processor of the online data fusion system, the first answer of the answers in response to the query.

2. The method of claim 1 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct further based upon a source quality of the first probed source of the at least two probed sources.

3. The method of claim 2 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct further based upon a coverage of the first probed source of the at least two probed sources.

4. The method of claim 1 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct further based upon a coverage of the first probed source of the at least two probed sources.

5. The method of claim 1 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct further based upon an answer expected from the additional source that is unprobed considering a copying relationship between the additional source and the first probed source of the at least two probed sources.

6. The method of claim 1 , further comprising receiving, using the online data fusion system, an ordered list of sources specifying an order in which to probe the at least two probed sources.

7. The method of claim 1 , wherein terminating probing without probing the additional source when the online data fusion system gains enough confidence that probing the additional source is unlikely to change the answers comprises terminating probing without probing the additional source when a termination condition is satisfied.

8. A computer storage medium comprising computer-executable instructions which, when executed by a computer, cause the computer to perform operations comprising:
receiving, in response to probing at least two probed sources, answers to a query from the at least two probed sources;
computing a probability that each answer of the answers is correct based, at least in part, upon a copying relationship between at least two of the at least two probed sources, wherein computing the probability that each answer of the answers is correct comprises
computing an expected probability, a maximum probability, and a minimum probability that each answer of the answers is correct, and
refreshing the expected probability, the maximum probability, and the minimum probability of a first answer of the answers from a first probed source of the at least two probed sources based, at least in part, on a second answer received from a second probed source of the at least two probed sources as the second answer is received from the second probed source of the at least two probed sources;

when the online data fusion system gains enough confidence that, based upon the probability that each answer of the answers is correct, probing an additional source is unlikely to change the first answer, terminating probing without probing the additional source; and
providing the first answer of the answers in response to the query.

9. The computer storage medium of claim 8 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct further based upon at least one of the following:
a source quality of the first probed source of the at least two probed sources; or
a coverage of the first probed source of the at least two probed sources.

10. The computer storage medium of claim 8 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct based upon an answer expected from the additional source that is unprobed considering a copying relationship between the additional source and the first probed source of the at least two probed sources.

11. The computer storage medium of claim 8 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct without knowledge of a quality of a probed source of the at least two probed sources.

12. The computer storage medium of claim 8 , wherein the operations further comprise receiving an ordered list of sources specifying an order in which to probe the at least two probed sources.

13. The computer storage medium of claim 8 , wherein terminating probing without probing the additional source when the online data fusion system gains enough confidence that probing the additional source is unlikely to change the answers comprises terminating probing without probing the additional source when a termination condition is satisfied.

14. The computer storage medium of claim 8 , wherein computing the probability that each answer of the answers is correct comprises computing the probability that each answer of the answers is correct further based upon a source quality of the first probed source of the at least two probed sources and a coverage of the first probed source of the at least two probed sources.

15. An online data fusion system comprising:
a processor; and
a memory that stores instructions which, when executed by the processor, cause the processor to perform operations comprising
receiving, in response to probing at least two probed sources, answers to a query from the at least two probed sources,
computing a probability that each answer of the answer is correct based, at least in part, upon a copying relationship between at least two of the at least two probed sources, wherein computing the probability that each answer of the answers is correct comprises
computing an expected probability, a maximum probability, and a minimum probability that each answer of the answers is correct, and
refreshing the expected probability, the maximum probability, and the minimum probability of a first answer of the answers from a first probed source of the at least two probed sources based, at least in part, on a second answer received from a second probed source o

Assignees

• Srivastava Divesh
• Dong Xin
• Liu Xuan
• At & T Ip I Lp

Inventors

• Srivastava Divesh
• Dong Xin
• Liu Xuan

Classifications

• G06F16/33 Primary

  Querying · CPC title

• G06F17/30634 Primary

  Physics · mapped topic