Learning apparatus and method of intelligent system

• US 8,356,002 B2
• Filed: 05/19/2008
• Issued: 01/15/2013
• Est. Priority Date: 09/04/2007
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Kim, Rockwon, Kim, Hyoungsun, Moon, Aekyeung, Kang, Taegun, Cho, Hyunkyu
• Primary Examiner(s): Chen, Alan
• Assistant Examiner(s): Kim, David H
• Application Number:
  US12/123,039
• Publication Number:
  US 20090063388A1
• Time in Patent Office: 1,702 Days
• Field of Search: 706/47, 704/275
• US Class Current: 706/47
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A learning apparatus and method of an intelligent system according to the present invention relates to learning of an intelligent system that automatically controls a system on the basis of information about events generated in an external environment or services input by a user. In the learning, it is determined whether rules extracted from event data or service data are the same as rules already stored. Then, if it is determined that the rules are new rules, queries about the confidence of new rules are generated and provided to a user. The rules are updated on the basis of user'"'"'s answers. Accordingly, it is possible to accurately grasp users'"'"' patterns by using very small amount of observation data as compared to an existing method, and to more easily sense variable patterns. Therefore, it is possible to obtain a more intelligent learning in the intelligent system.

Claims

1. A learning apparatus of an intelligent system that automatically controls a system on the basis of information about events generated in an external environment or services input by a user, the learning apparatus comprising:

a rule management device determining whether rules extracted from event data or service data are the same as rules already stored in a rules disc storage, generating queries about the confidence of new rules and providing the queries to the user if the rules are new rules, and updating the rules on the basis of user'"'"'s answers; and

the rules disc rule storage that stores the rules in a confidence region or a non-confidence region depending on the confidence of rules input by the user.

2. The learning apparatus according to claim 1,

wherein the rule management device includes:

a translation table used to translate services and events, which are included in rules extracted from a learning engine, into strings;

a query generation module that generates user queries by using the translation table;

a speech module that delivers the user queries to a user by TTS (Text to Speech); and

a response recognition module that receives answers from the user.

3. The learning apparatus according to claim 1, further comprising:

a logging device that stores sensed event information or log information about service requests;

an event analyzing device that analyzes event or service information input from the logging device, classifies the event or service information into data to be learned, and preprocesses the data;

a learning engine that implements a learning algorithm by using the data preprocessed in the event analyzing device so as to search users'"'"' patterns included in the preprocessed data; and

a rule extracting device that converts the results searched by the learning engine into rule formats, and extracts rules.

4. The learning apparatus according to claim 3,

wherein the event analyzing device includes:

a table that stores meta information about an event source or meta information about services implemented by a service terminal; and

a log binding module that searches service information related to event information and relates service information to event information.

5. The learning apparatus according to claim 1, further comprising:

a rule engine that processes logical conclusion and queries about the stored rules; and

a rule trigger that searches services corresponding to event or service information that is currently input from rules already stored, and provides corresponding services or related services to the user.

6. A learning method of an intelligent system that automatically controls a system on the basis of information about events generated in an external environment or services input by a user, the learning method comprising:

determining whether rules extracted from event data or service data are the same as rules already stored;

generating queries about the confidence of new rules and providing the queries to a user if the rules are the new rules;

updating the rules on the basis of user'"'"'s answers; and

storing the rules in a confidence region or a non-confidence region depending on the confidence of rules input by the user.

7. The learning method according to claim 6, further comprising:

generating user queries by using a translation table, the translation table being used to translate services and events, which are included in rules extracted from a learning engine, into strings;

delivering the generated user queries to a user by TTS (Text to Speech); and

receiving answers from the user.

8. The learning method according to claim 6, further comprising:

storing sensed event information or log information about service requests;

analyzing event or service information input from the storing of the information, classifying the event or service information into data to be learned, and preprocessing the data; and

implementing a learning algorithm by using the preprocessed data so as to search users'"'"' patterns included in the preprocessed data.

9. The learning method according to claim 6, further comprising:

processing logical conclusion and queries about the stored rules; and

searching services corresponding to event or service information that is currently input from rules already stored, and providing corresponding services or related services to the user.