LEARNING A DOCUMENT RANKING FUNCTION USING FIDELITY-BASED ERROR MEASUREMENTS

• US 20080027912A1
• Filed: 07/31/2006
• Published: 01/31/2008
• Est. Priority Date: 07/31/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ma, Wei-Ying, Liu, Tie-Yan, Tsai, Ming-Feng
• Granted Patent: US 7,805,438 B2
• Time in Patent Office: Days
• US Class Current: 707/3
• CPC Class Codes: G06F 16/3346 using probabilistic model

Abstract

A method and system for generating a ranking function using a fidelity-based loss between a target probability and a model probability for a pair of documents is provided. A fidelity ranking system generates a fidelity ranking function that ranks the relevance of documents to queries. The fidelity ranking system operates to minimize a fidelity loss between pairs of documents of training data. The fidelity loss may be derived from “fidelity” as used in the field of quantum physics. The fidelity ranking system may use a learning technique in conjunction with a fidelity loss when generating the ranking function. After the fidelity ranking system generates the fidelity ranking function, it uses the fidelity ranking function to rank the relevance of documents to queries.

Claims

1. A method in a computer system for determining loss between a target probability and a model probability for a pair of documents, the model probability being generated by a ranking function that ranks documents, the method comprising:

applying the ranking function to the pair of documents to provide rankings of the documents;

calculating a model probability from the rankings of the documents; and

calculating a loss between the calculated model probability and the target probability, the loss varying between 0 and 1 and the loss being 0 when the calculated model probability is the same as the target probability.

2. The method of claim 1 wherein the calculated loss is a fidelity loss.

3. The method of claim 2 wherein the fidelity-based loss is represented by the following equation:

F_ij=1−

(√

{square root over (P_ij*·

P_ij)}+√

{square root over ((1−

P_ij*)·

(1−

P_ij))}{square root over ((1−

P_ij*)·

(1−

P_ij))})

where F_ij represents the fidelity loss, P_ij* represents the target probability for documents i and j, and P_ij represents the calculated model probability for documents i and j.

4. The method of claim 1 wherein the applying of the ranking function, the calculating of a model probability, and the calculating of a loss are performed when generating a ranking function.

5. The method of claim 4 wherein the generating of a ranking function uses an adaptive boosting technique.

6. The method of claim 4 wherein the generating of a ranking function uses a neural network technique.

7. The method of claim 4 wherein the generating of a ranking function uses a support vector machine technique.

8. The method of claim 1 wherein the calculating of a model probability applies a logistic function to the ranking of the documents.

9. The method of claim 8 wherein the logistic function is represented by the following equation:

$P_{\mathrm{ij}}=\frac{{\mathrm{<br><br>}}^{o_{\mathrm{ij}}}}{1+{\mathrm{<br><br>}}^{o_{\mathrm{ij}}}}$

where P_ij represents the probability that document i is ranked higher than document j and o_ij represents the difference between the outputs of the fidelity ranking function for document i and document j as represented by f(d_i)−

f(d_j) with f(d_i) being the output of the fidelity ranking function for document i.

10. A computer system for generating a ranking function for documents, the ranking function indicating a ranking of documents based on relevance of the documents to a query, the system comprising:

a component that provides features of documents and indications of target probabilities of relative rankings of the relevance of pairs of documents to queries;

a component that calculates a fidelity loss between a model probability and a target probability for a pair of documents, the probabilities indicating a probability of relative ranking of the documents of the pair; and

a component that generates the ranking function by operating to minimize the fidelity loss between the model probabilities derived from the ranking of documents and the target probabilities.

11. The computer system of claim 10 wherein the fidelity loss varies between 0 and 1 and the fidelity loss is 0 when the model probability is the same as the target probability.

12. The computer system of claim 11 wherein the fidelity-based loss is represented by the following equation:

F_ij=1−

(√

{square root over (P_ij*·

P_ij)}+√

{square root over ((1−

P_ij*)·

(1−

P_ij))}{square root over ((1−

P_ij*)·

(1−

P_ij))})

where F_ij represents the fidelity loss, P_ij* represents the target probability for documents i and j, and P_ij represents the calculated model probability for documents i and j.

13. The computer system of claim 10 wherein the model probability is derived by applying a logistic function to the ranking of the documents.

14. The computer system of claim 13 wherein the logistic function is represented by the following equation:

$P_{\mathrm{ij}}=\frac{{\mathrm{<br><br>}}^{o_{\mathrm{ij}}}}{1+{\mathrm{<br><br>}}^{o_{\mathrm{ij}}}}$

where P_ij represents the probability that document i is ranked higher than document j and o_ij represents the difference between the outputs of the fidelity ranking function for document i and document j as represented by f(d_i)−

f(d_j) with f(d_i) being the output of the fidelity ranking function for document i.

15. The computer system of claim 10 wherein the component that generates a ranking function uses an adaptive boosting technique.

16. The computer system of claim 10 wherein the component that generates a ranking function uses a neural network technique.

17. The computer system of claim 10 wherein the component that generates a ranking function uses a support vector machine technique.

18. A computer system for ranking relevance of documents to queries, the system comprising:

a component that receives a query;

a component that identifies documents that match the query; and

a component that applies a fidelity ranking function to the documents that match the query to rank relevance of the documents to the queries, the fidelity ranking function being generated to minimize a fidelity loss between model probabilities generated by the fidelity ranking function and target probabilities.

19. The computer system of claim 18 including a component to extract features of the documents wherein the fidelity ranking function generates a ranking based on features of the documents.

20. The computer system of claim 18 wherein the fidelity loss is represented by the following equation:

F_ij=1−

(√

{square root over (P_ij*·

P_ij)}+√

{square root over ((1−

P_ij*)·

(1−

P_ij))}{square root over ((1−

P_ij*)·

(1−

P_ij))})

where F_ij represents the fidelity loss, P_ij* represents the target probability for documents i and j, and P_ij represents the calculated model probability for documents i and j.