Method and apparatus for open data collection

• US 6,879,967 B1
• Filed: 03/24/2000
• Issued: 04/12/2005
• Est. Priority Date: 03/24/2000
• Status: Expired due to Term

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Stork, David G.
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/534,950
• Time in Patent Office: 1,845 Days
• Field of Search: 706/10, 706/12, 706/45, 706/50, 706/16, 707/6, 707/1
• US Class Current: 706/12
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

A method and apparatus for open data collection is provided. The method of machine learning comprises setting up a system for learning, the system having certain goals. The method further comprising presenting queries to non-expert netizens over a network, the netizens participating in the training process. Then, based on responses to the queries, the method continually updating the system and refining the queries.

Claims

1. A method of machine learning using a training process to train a learning system, the method comprising:

presenting queries to non-expert netizens over a network, the netizens participating in the training process; and

continually updating the system and refining the queries based on responses to the queries provided by the netizens,

wherein the queries are multiple choice queries.

2. The method of claim 1, wherein the system has certain goals including accumulating data.

3. The method of claim 2, wherein at least one goal comprises a goal selected from among the following:

handwriting recognition, voice recognition, building a database of queries to recognize an object, building a database of common sense.

4. The method of claim 2, wherein the goals of the system evolve as the system is updated.

5. The method of claim 4, wherein the goals comprise a plurality of intermediate goals, that change in response to the responses while approaching a final goal.

6. The method of claim 5, wherein one of the plurality of intermediate goals is to recognize a certain letter of the alphabet in handwriting.

7. The method of claim 5, wherein one of the plurality of intermediate goals is to recognize a sound corresponding to a certain set of letters, in context.

8. The method of claim 1, further comprising providing access to a domain expert to resolve conflicts between the responses of netizens, if a conflict arises.

9. The method of claim 1, wherein setting up the system comprises:

implementing a plurality of rules for presenting questions;

implementing an architecture for interacting with the netizens to enable netizens to access the system; and

generating a database for storing the responses.

10. The method of claim 9, further comprising:

evaluating a reliability rating for each of the netizens; and

weighting the response of each of the netizens according to the reliability rating.

11. A system coupled to a network to present queries to and receive responses from a plurality of netizens over the network, the system comprising:

a user interface to present the queries and receiving the responses;

a data aggregation logic to organize the responses;

a query formulation logic to formulate a next query based on the plurality of responses to the last query; and

reliability evaluation logic to weight each response according to a reliability of the netizen providing the response.

12. The system of claim 11, further comprising:

conflict resolution logic to resolve conflicts between responses provided by the netizens using domain experts.

13. A method of data aggregation over a network comprising:

presenting a question to a plurality of participants over a network;

receiving responses to the question;

analyzing the plurality of responses to the question from the plurality of participants;

formulating a next question based on the plurality of responses; and

presenting the next question to the plurality of participants.

14. The method of claim 13, further comprising:

resolving a conflict between the plurality responses provided by the netizens using domain experts, if the conflict arises.

15. The method of claim 13, further comprising:

evaluating a reliability rating for each of the netizens; and

weighting the response of each of the netizens according to the reliability rating.

16. A method of interacting with a user comprising:

presenting a query to the user over a network;

receiving a response to the query from the user, the response transmitted to a learning system; and

informing the user of a result generated based on the response to the query, such that the user is rewarded by being informed of the content and state of data being gathered based on the response.

17. A machine readable medium having stored thereon data representing sequences of instructions, which when executed by a computer system, cause said computer system to perform the operations of:

presenting multiple choice queries to non-expert netizens over a network, the netizens participating in a training process of a learning system; and

continually updating the learning system and refining the multiple choice queries based on responses to the queries provided by the netizens.

18. The machine readable medium of claim 17, wherein the system includes a plurality of goals, and one of the goals is to accumulate data.

19. The machine readable medium of claim 17, further comprising:

rewarding netizens for their participation in the training process.

20. A system for implementing a training process comprising:

a means for presenting queries to and receiving responses from non-expert netizens over a network, the netizens participating in the training process;

a means for continually updating the system and refining the queries based on the responses to the queries provided by the netizens; and

a means for rewarding the netizens for participation in training the system.

21. The system for training of claim 20, further comprising:

a means for storing the responses of the netizens; and

a means for weighting the responses of each netizens based on a reliability of the netizen.