Similarity evaluation method, similarity evaluation program and similarity evaluation apparatus

• US 20030065510A1
• Filed: 03/28/2002
• Published: 04/03/2003
• Est. Priority Date: 09/28/2001
• Status: Abandoned Application

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Sato, Makihiko
• Application Number:
  US10/107,204
• Publication Number:
  US 20030065510A1
• Time in Patent Office: Days
• US Class Current: 704/239
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04

Abstract

A similarity evaluation program capable of determining similarity between probability models at a high speed (with little calculation) is disclosed. The similarity evaluation program is implemented on an apparatus such as a computer for evaluating similarity between a pair of probability model information each including a plurality of probability information constituted by a plurality of types of data, and this apparatus is provided with a dynamic programming operation unit for performing arithmetic processing based on dynamic programming techniques using a similarity value indicating similarity between probability information included in one of the pair probability model information and probability information included in the other of the pair of probability model information as an indicator for selecting a path.

Claims

1. A similarity evaluation method for evaluating similarity between a pair of probability model information each including a plurality of items of probability information constituted by a plurality of probability data, including:

a similarity value calculating step of calculating a similarity value indicating similarity between a pair of probability information based on probability information included in one of the pair of probability model information and probability information included in the other of the pair of probability model information; and

an evaluating step of carrying out arithmetic processing based on dynamic programming techniques employing a similarity value calculated in the similarity value calculating step as an indicator value for path selection, whereby evaluating similarity of the pair of probability model information.

2. The similarity evaluation method according to claim 1, wherein

the probability model information is probability model information relating to a Hidden Markov Model, and

the similarity value calculating step involves calculating the similarity value based on a plurality of emission probability data in the probability information included the one of the pair of probability information, and a plurality of emission probability data in the probability information included in the other of the pair of probability information.

3. The similarity evaluation method according to claim 1, wherein

the probability model information is probability model information relating to a Hidden Markov Model, and

the similarity value calculating step involves calculating the similarity value based on a plurality of emission probability data and a plurality of transition probability data in the probability information included in the one of the pair of probability information, and a plurality of emission probability data and a plurality of transition probability data in the probability information included in the other of the pair of probability information.

4. The similarity evaluation method according to claim 3, wherein

the similarity value calculating step involves calculating the similarity value by multiplying the cosine squared of an angle made by an emission probability vector constituted by a plurality of emission probability data in the probability information included in the one of the pair of probability information and an emission probability vector constituted by a plurality of emission probability data in the probability information included in the other of the pair of probability information with the cosine squared of an angle made by a transition probability vector constituted by the plurality of transition probability data in the probability information included in the one of the pair of probability information and a transition probability vector constituted by the plurality of transition probability data in the probability information included in the other of the pair of probability information.

5. The similarity evaluation method of any one according to claim 1 to claim 4, further including a step of outputting information indicating the relationship of similarity between pairs of probability information include in the pair of probability model information based on the results of arithmetic processing carried out in the evaluating step.

6. A similarity evaluation program for making a computer to perform a process of evaluating similarity between a pair of probability model information each including a plurality of items of probability information constituted by a plurality of probability data, said process comprising:

a similarity value calculating step of calculating a similarity value indicating similarity between a pair of probability information based on probability information included in one of the pair of probability model information and probability information included in the other of the pair of probability model information; and

an evaluating step of carrying out arithmetic processing based on dynamic programming techniques employing a similarity value calculated in the similarity value calculating step as an indicator value for path selection, whereby evaluating similarity of the pair of probability model information.

7. The similarity evaluation method according to claim 6, wherein

the probability model information is probability model information relating to a Hidden Markov Model, and

the similarity value calculating step involves calculating the similarity value based on a plurality of emission probability data in the probability information included the one of the pair of probability information, and a plurality of emission probability data in the probability information included in the other of the pair of probability information.

8. The similarity evaluation method according to claim 6, wherein

the probability model information is probability model information relating to a Hidden Markov Model, and

the similarity value calculating step involves calculating the similarity value based on a plurality of emission probability data and a plurality of transition probability data in the probability information included in the one of the pair of probability information, and a plurality of emission probability data and a plurality of transition probability data in the probability information included in the other of the pair of probability information.

9. The similarity evaluation method according to claim 8, wherein

the similarity value calculating step involves calculating the similarity value by multiplying the cosine squared of an angle made by an emission probability vector constituted by a plurality of emission probability data in the probability information included in the one of the pair of probability information and an emission probability vector constituted by a plurality of emission probability data in the probability information included in the other of the pair of probability information with the cosine squared of an angle made by a transition probability vector constituted by the plurality of transition probability data in the probability information included in the one of the pair of probability information and a transition probability vector constituted by the plurality of transition probability data in the probability information included in the other of the pair of probability information.

10. The similarity evaluation method according to any one of claim 6 to claim 9, further including a step of outputting information indicating the relationship of similarity between pairs of probability information include in the pair of probability model information based on the results of arithmetic processing carried out in the evaluating step.

11. A similarity evaluation apparatus for evaluating similarity between a pair of probability model information each including a plurality of items of probability information constituted by a plurality of probability data, comprising:

a similarity value calculating part for calculating a similarity value indicating similarity between a pair of probability information based on probability information included in one of the pair of probability model information and probability information included in the other of the pair of probability model information; and

an evaluating part for carrying out arithmetic processing based on dynamic programming techniques employing a similarity value calculated by the similarity value calculating part as an indicator value for path selection, whereby evaluating similarity of the pair of probability model information.

12. The similarity evaluation apparatus according to claim 11, wherein

the probability model information is probability model information relating to a Hidden Markov Model, and

the similarity value calculating part calculates the similarity value based on a plurality of emission probability data in the probability information included the one of the pair of probability information, and a plurality of emission probability data in the probability information included in the other of the pair of probability information.

13. The similarity evaluation apparatus according to claim 11, wherein

the probability model information is probability model information relating to a Hidden Markov Model, and

the similarity value calculating part calculates the similarity value based on a plurality of emission probability data and a plurality of transition probability data in the probability information included in the one of the pair of probability information, and a plurality of emission probability data and a plurality of transition probability data in the probability information included in the other of the pair of probability information.

14. The similarity evaluation apparatus according to claim 13, wherein

the similarity value calculating part calculates the similarity value by multiplying the cosine squared of an angle made by an emission probability vector constituted by a plurality of emission probability data in the probability information included in the one of the pair of probability information and an emission probability vector constituted by a plurality of emission probability data in the probability information included in the other of the pair of probability information with the cosine squared of an angle made by a transition probability vector constituted by the plurality of transition probability data in the probability information included in the one of the pair of probability information and a transition probability vector constituted by the plurality of transition probability data in the probability information included in the other of the pair of probability information.

15. The similarity evaluation apparatus according to any one of claim 11 to claim 14, further comprising a part for outputting information indicating the relationship of similarity between pairs of probability information include in the pair of probability model information based on the results of arithmetic processing carried out by the evaluating part.