Information processing apparatus, method and program

1. An information processing apparatus, which creates a tree structure used by a recognition apparatus which recognizes specific information using the tree structure, comprising:

memory means which stores data including the information to be recognized and a label identifying the information;

a plurality of recognition devices which recognize the information and output a high score value when the data including the information is in a class corresponding to a recognition device into which the data is input; and

grouping means which performs grouping of the plurality of recognition devices using a score distribution obtained when each of the data is input into each of the plurality of recognition devices, the grouping means merging classes corresponding to all of the recognition devices placed in a group such that all of the data for each of the classes corresponding to each of the recognition devices in a group is placed into a single class to create grouped devices, the grouping means calculating a distance of the score distribution for each of the plurality of recognition devices and grouping all of the plurality of recognition devices having scores within a threshold.

2. The information processing apparatus according to claim 1 , wherein the grouping means calculates a distance of the score distribution and performs grouping by clustering the distances.

3. The information processing apparatus according to claim 2 , wherein the calculation of the distance of the score distribution uses chi-squared distance.

4. The information processing apparatus according to claim 2 , wherein the calculation of the distance of the score distribution uses KL-divergence.

5. The information processing apparatus according to claim 2 , wherein the clustering uses spectral clustering.

6. The information processing apparatus according to claim 1 , wherein each recognition device is a discrimination device which has learned in advance.

7. The information processing apparatus according to claim 1 , wherein the information is a specific object or a specific voice, and the data is image data, video data or voice data.

8. An information processing method of an information processing apparatus, which is an information processing apparatus which creates a tree structure used by a recognition apparatus which recognizes specific information using the tree structure, and which at least is provided with memory means which stores data including the information to be recognized and a label identifying the information, comprising:

performing grouping of a plurality of recognition devices using a score distribution obtained when each of the data read from the memory means is input into each of the plurality of recognition devices which recognize the information and which output a high score value when the data including the information is in a class corresponding to a recognition device into which the data is input, the grouping including merging classes corresponding to all of the recognition devices placed in a group such that all of the data for each of the classes corresponding to each of the recognition devices in a group is placed into a single class, the grouping including calculating a distance of the score distribution for each of the plurality of recognition devices and grouping all of the plurality of recognition devices having scores within a threshold.

9. A non-transitory computer readable medium encoded with a computer-readable program which executes processing in an information processing apparatus which creates a tree structure used by a recognition apparatus recognizing specific information using the tree structure, and which at least is provided with memory means storing data including the information to be recognized and a label identifying the information, comprising:

performing grouping of a plurality of recognition devices using a score distribution obtained when each of the data read from the memory means is input into each of the plurality of recognition devices which recognize the information and output a high score value when the data including the information is in a class corresponding to a recognition device into which the data is input, the grouping including merging classes corresponding to all of the recognition devices placed in a group such that all of the data for each of the classes corresponding to each of the recognition devices in a group is placed into a single class, the grouping including calculating a distance of the score distribution for each of the plurality of recognition devices and grouping all of the plurality of recognition devices having scores within a threshold.

10. An information processing apparatus, which creates a tree structure used by a recognition apparatus which recognizes specific information using the tree structure, comprising:

a memory unit which stores data including the information to be recognized and a label identifying the information;

a plurality of recognition devices which recognize the information and output a high score value when the data including the information is in a class corresponding to a recognition device into which the data is input; and

a grouping unit which performs grouping of the plurality of recognition devices using a score distribution obtained when each of the data is input into each of the plurality of recognition devices, the grouping unit merging classes corresponding to all of the recognition devices placed in a group such that all of the data for each of the classes corresponding to each of the recognition devices in a group is placed into a single class, the grouping unit calculating a distance of the score distribution for each of the plurality of recognition devices and grouping all of the plurality of recognition devices having scores within a threshold.