Rule based expert system for consumer preference

• US 6,741,975 B1
• Filed: 08/31/2000
• Issued: 05/25/2004
• Est. Priority Date: 09/01/1999
• Status: Expired due to Term

• Current Assignee: NCR Corporation
• Original Assignee: NCR Corporation
• Inventors: Mackay, Robin, Nakisa, Ramin C.
• Primary Examiner(s): Davis, George B.
• Application Number:
  US09/651,981
• Time in Patent Office: 1,363 Days
• Field of Search: 706/47, 706/48, 706/45
• US Class Current: 706/47
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06Q 40/02 Banking, e.g. interest calculation or account maintenance credit or loans G06Q40/03

Abstract

An expert system and a method of providing automated advice are described. The system is regularly updated by advice (or diagnosis, recommendation etc.) given by practitioners in the relevant field. The combination of the underlying facts and the consequent (human) advice is used to update a ruleset that is then used to provide automated advice. An example of financial advice is given. A database contains the details for the customers of a financial institution together with advice and recommendations given previously by the institution'"'"'s human advisors. This database is used to derive a ruleset which is then applied to a subsequent customer'"'"'s details in an automated manner, possibly at the user'"'"'s own PC via the World Wide Web. Frequent updating using data from the human advisors'"'"' work means that the database (and hence the ruleset) are kept up to date. Consistent advice can thus be provided with minimum human interaction.

Claims

1. A knowledge-based system for providing a recommendation tailored to a consumer, the system comprising:

a knowledge base containing historical data;

rule extraction means for extracting a ruleset from the knowledge base;

a rules database for holding the ruleset;

codifying means for codifying the requirements of the consumer; and

recommendation means for applying the ruleset to the codified consumer requirements and generating the recommendation accordingly;

explanation means for explaining to the consumer at least one reason for the recommendation; and

means for using XML (Extensible Markup Language) to define consumers and possible/actual recommendations.

2. A system according to claim 1, wherein the recommendation means is configured to provide the recommendation as a document based on an XML DTD (document type definition).

3. A system according to claim 1, wherein the explanation means is associated with the rule extraction means for locating the rules that govern decisions reached by the recommendation means in making the recommendation, and is configured to explain to the consumer the rules on which the recommendation was based.

4. A system according to claim 1, wherein the explanation means is configured to express the explanation substantially in a natural language.

5. A system according to claim 1, wherein the rule extraction means runs Lozowski'"'"'s algorithm.

6. A system according to claim 1, wherein the recommendation means is configured to suggest a plurality of alternative recommendations.

7. A system according to claim 6, wherein the recommendation means is configured to rank recommendations by their suitability.

8. A system according to claim 1, wherein the codifying means includes means for codifying the perceived needs of the consumer and means for codifying the circumstances of the consumer.

9. A system according to claim 1, further comprising storage means for storing a consumer'"'"'s details for later recall.

10. A system according to claim 9, wherein the storage means stores the consumer'"'"'s details for later recall in providing a future recommendation.

11. A system according to claim 10, wherein the storage means stores the consumer'"'"'s details for later recall in completing a recommendation, and is associated with means for suspending input of consumer requirements.

12. A system according to claim 1, further comprising communications means for obtaining advice from a remote human adviser.

13. A system according to claim 12, wherein the communications means comprises a video conference link between the consumer and the adviser.

14. A system according to claim 1, wherein the system comprises a distributed system.

15. A system according to claim 14, further comprising a server holding the ruleset and data on consumers and on possible recommendations, a consumer terminal providing an online interface with the server, and a plurality of expert terminals operable by the plurality of human experts, each expert terminal including means for storing recommendations made by an expert and the requirements of consumers that that expert has advised, and means for providing that stored data to the server for use in updating the ruleset.

16. A system according to claim 15, wherein the consumer terminal includes a home PC or a kiosk, booth, ATM or other terminal in a financial advice establishment.

17. A system according to claim 16, further comprising an application server enabling consumer terminals and expert terminals to interact with the server online, the application server providing an online interface to the server for the consumer terminals and the expert terminals.

18. A system according to claim 17, wherein the application server runs server-side web applications, a first web application responding to calls from a consumer website, and a second web application allowing access to the server by authorized expert terminals.

19. A system according to claim 18, wherein the server-side web applications include Java Servlets.

20. A system according to claim 1, wherein the ruleset and data on consumers and on possible recommendations are stored at the server as XML (Extensible Markup Language) documents.

21. A system according to claim 1, wherein the recommendation means includes an agent that applies the rules to the codified consumer requirements to generate the recommendation.

22. A system according to claim 21, wherein the agent includes questioning means for obtaining the consumer'"'"'s requirements, the questioning means asking a sequence of questions and including means for adapting later questions in the sequence in accordance with answers given to earlier questions in the sequence.

23. A system according to claim 1, further comprising means for updating the knowledge base with changing details of recommendations that can be made.

24. A method of operating a knowledge-based system to provide a recommendation tailored to a consumer, the method comprising:

extracting rules from a knowledge base; and

applying the extracted rules to codified consumer requirements to generate the recommendation accordingly and explaining to the consumer at least one reason for the recommendation, wherein XML (Extensible Markup Language) is used to define consumers and possible/actual recommendations.

25. A method according to claim 24, wherein the recommendation is provided to the consumer as a document based on an XML DTD (document type definition).

26. A method according to claim 24, further comprising:

explaining to the consumer the rules on which the recommendation was based.

27. A method according to claim 24, further comprising:

expressing the explanation substantially in a natural language.

28. A method according to claim 24, wherein Lozowski'"'"'s algorithm is applied to extract rules from the knowledge base.

29. A method according to claim 24, further comprising suggesting a plurality of alternative recommendations.

30. A method according to claim 29, further comprising ranking recommendations by their suitability.

31. A method according to claim 24, further comprising updating the knowledge base with changing details of recommendations that can be made.

32. A method according to claim 24, wherein the codified consumer requirements include the perceived needs of the consumer and the circumstances of the consumer.

33. A method according to claim 24, further comprising storing a consumer'"'"'s details for later recall.

34. A method according to claim 33, wherein the consumer'"'"'s details are stored and later recalled when providing a future recommendation.

35. A method according to claim 33, wherein the consumer'"'"'s details are stored and later recalled when completing a recommendation, input of consumer requirements being suspended in the meantime.

36. A method according to claim 24, wherein the consumer obtains advice from a remote human adviser.

37. A method according to claim 36, wherein the advice is obtained over a video conference link.

38. A method according to claim 37, wherein the consumer enters consumer requirements via a consumer terminal being a home PC or a kiosk, booth, ATM or other terminal in a financial advice establishment.

39. A method according to claim 38, wherein the consumer enters consumer requirements in response to a sequence of questions, and wherein later questions in the sequence are adapted in accordance with answers given to earlier questions in the sequence.