Context-Based Suggestions Mechanism and Adaptive Push Mechanism for Natural Language Systems

• US 20090144248A1
• Filed: 11/26/2008
• Published: 06/04/2009
• Est. Priority Date: 12/20/2001
• Status: Active Grant

• Current Assignee: Sybase 365 Incorporated (SAP SE)
• Original Assignee: Sybase 365 Incorporated (SAP SE)
• Inventors: Treadgold, Nicholas K., Hodjat, Babak
• Granted Patent: US 8,036,877 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

Natural language interface to a back-end application, incorporating synonyms, suggestions, and proposals. Roughly described, synonyms are automatically added to user input to enhance the natural language interpretation, whereas suggestions and proposals are offered to the user in an interaction, usually after an interpretation of prior user input. Suggestions and synonyms can be learned from user input, whereas proposals are programmed by a third party. The selection of synonyms, suggestions, and proposals for use with particular user input can be context-based, and rewards-based reinforcement can be used to better focus suggestions and proposals on the characteristics of the particular user.

Claims

1. A method for user input to a back-end application, comprising the steps of:

receiving first user input in the form of one or more tokens;

querying a synonyms database for synonyms associated with said tokens;

attempting a natural language interpretation of said first user input in combination with the synonyms associated with said tokens;

forwarding a result of said natural language interpretation and said first user input to a suggestions and proposals database;

identifying in said suggestions and proposals database, additional potential input in dependence upon tokens in said first user input; and

offering said additional potential input to a user.

2. A method according to claim 1, wherein said step of attempting a natural language interpretation of said first user input at least partially fails.

3. A method according to claim 2, wherein said step of identifying additional potential input in dependence upon tokens in said first user input, comprises the step of identifying said additional potential input in dependence upon only those tokens in said first user input of which said step of attempting a natural language interpretation failed.

4. A method according to claim 2, wherein said step of attempting a natural language interpretation produces an interpretation of a token in said first user input, but with a confidence level that is below a predetermined minimum acceptable confidence level.

5. A method according to claim 1, wherein additional potential input entries in said database each associate a target token sequence with at least one potential additional token sequence, and wherein said step of identifying additional potential input comprises the steps of:

comparing tokens in said user input to said target token sequences; and

selecting a set of at least one of said additional potential input entries in response to said step of comparing.

6. A method according to claim 1, wherein said step of attempting a natural language interpretation produces an interpretation of a token in said first user input,

wherein additional potential input entries in said database each associate a target token sequence with at least one potential additional token sequence,

and wherein said step of identifying additional potential input comprises the steps of:

comparing tokens in said interpretation to said target token sequences; and

selecting a set of at least one of said additional potential input entries in response to said step of comparing.

7. A method according to claim 1, wherein said step of attempting a natural language interpretation produces an interpretation of a token in said first user input,

wherein additional potential input entries in said database each associate a target token sequence with at least one potential additional token sequence, said database also including a respective importance value for each of said associations,

and wherein said step of identifying additional potential input comprises the steps of:

comparing to said target token sequences, tokens in at least one of said user input and said interpretation; and

selecting a set of at least one of said additional potential input entries in response to said step of comparing and in dependence upon said importance values.

8. A method according to claim 7, wherein said step of selecting a set of at least one of said additional potential input entries comprises the steps of:

for each particular one of said associations, calculating a confidence score given by the degree to which said target token sequences of the particular association match the tokens in said at least one of said user input and said interpretation, scaled by the importance value of the particular association; and

selecting the additional potential input entries from only those of said associations having a confidence score exceeding a predetermined minimum acceptable confidence score.

9. A method according to claim 7, wherein said step of selecting a set of at least one of said additional potential input entries comprises the steps of:

for each particular one of said associations, calculating a confidence score given by the degree to which said target token sequences of the particular association match the tokens in said at least one of said user input and said interpretation, scaled by the importance value of the particular association; and

selecting the additional potential input entries of only those of said associations having a confidence score in the top N confidence scores calculated in said step of calculating, N being a predetermined number.

10. A method according to claim 7, wherein said step of selecting a set of at least one of said additional potential input entries comprises the steps of:

for each particular one of said associations, calculating a context matching score given by the degree to which said target token sequences of the particular association match the tokens in said at least one of said user input and said interpretation; and

selecting the additional potential input entries in dependence upon both said context matching scores and said importance values,

further comprising the step of negatively reinforcing the importance value of a subject association in said database whose additional potential input is offered to but not accepted by said user, said negative reinforcement being more severe the greater the context matching score for said subject association.

11. A method according to claim 5, wherein said step of offering said additional potential input comprises the step of offering at least one of the additional potential input token sequences to said user from said set of additional potential input entries.

12. A method according to claim 5, wherein additional potential input entries in said database each further include display text in correspondence with each of said potential additional token sequences,

and wherein said step of offering said additional potential input comprises the step of forwarding toward said user the display text corresponding to at least one of the additional potential input token sequences in said set of additional potential input entries.

13. A method according to claim 5, wherein said database further includes an importance value in correspondence with each association of a potential additional token sequence with a target token sequence,

and wherein said step of comparing tokens in said user input to said target token sequences includes a step of calculating a confidence value of a given association in said database between a potential additional token sequence and a target token sequence, said confidence value being dependent upon both the importance value of the given association and a correlation between said tokens in said user input and the target token sequence of the given association.

14. A method according to claim 13, wherein said step of selecting comprises the step of comparing said confidence value to a predetermined minimum acceptable confidence value.

15. A method according to claim 13, further comprising the step of, in response to second user input following said step of offering, updating importance values in said database of at least one of the associations the additional potential input of which was offered to said user in said step of offering.

16. A method according to claim 15, wherein said second user input is not a follow-up

to said first user input,

and wherein said step of updating importance values in said database comprises the step of reducing the importance values of each of the associations the additional potential input of which were offered to said user in response to said first user input.

17. A method according to claim 15, wherein said second user input includes acceptance of the additional potential input offered in said step of offering, and wherein said step of updating importance values in said database comprises the steps of:

increasing the importance value of the association the additional potential input of which was accepted by said user; and

reducing the importance values of each of the associations the additional potential input of which were offered to said user in response to said first user input, other than the association the additional potential input of which was accepted by said user.

18. A method according to claim 1, further comprising the step of, in response to user acceptance of the additional potential input offered in said step of offering, attempting a natural language interpretation of second user input including said additional potential input.

19. A method according to claim 1, wherein said additional potential input comprises a proposal.

20. A method according to claim 1, further comprising the step of an operator other than said user preparing said database.

21-38. -38. (canceled)