Processing context-based inquiries for knowledge retrieval

• US 10,769,138 B2
• Filed: 06/13/2017
• Issued: 09/08/2020
• Est. Priority Date: 06/13/2017
• Status: Active Grant

• Current Assignee: Merative US LP
• Original Assignee: International Business Machines Corporation
• Inventors: Bagchi, Sugato, Barker, Kenneth J., Boguraev, Branimir K., Bornea, Mihaela A., Faulkner, Adam R., Li, Yanpeng, Patwardhan, Siddharth A., Rosenthal, Sara
• Primary Examiner(s): Yen, Syling
• Application Number:
  US15/621,623
• Publication Number:
  US 20180357272A1
• Time in Patent Office: 1,183 Days
• Field of Search: 707758
• CPC Class Codes: G06F 16/243 Natural language query formulation
• CPC Class Codes: G06F 16/24568 Data stream processing; Continuous queries
• CPC Class Codes: G06F 16/283 Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP

Abstract

Provided herein are techniques for processing a context-situated inquiry to provide results satisfying the inquiry. An inquiry and its supporting context are processed using natural language processing to determine an interpretation for the inquiry and context, and the interpretation is presented to receive corrective information for the interpretation. The corrective information is applied to the interpretation and content is retrieved based on the interpretation of inquiry and context to produce candidate results for the inquiry. Supplemental information is iteratively received to generate updated candidate results.

Claims

1. A computer program product for processing an inquiry to provide results satisfying the inquiry, the computer program product comprising one or more computer readable storage media collectively having computer readable program code embodied therewith, the computer readable program code executable by a processor to cause the processor to:

process a context using natural language processing to determine an interpretation for the context, wherein the context includes a description of a scenario pertaining to the inquiry and the interpretation for the context includes results of the natural language processing from one or more of parsing, entity detection, and relation detection of the context, and wherein the results of the natural language processing of the context include one or more factors each comprising a context portion including text extracted from the description of the scenario, and a type determined by the natural language processing for each factor comprising a description of the interpretation for the context portion corresponding to the factor;

process the inquiry using natural language processing to determine an interpretation for the inquiry based on the interpretation for the context;

present the interpretation for the inquiry to a user;

in response to an incorrect interpretation for the inquiry;

receive corrective information for the interpretation for the context from the user, wherein the corrective information for the interpretation for the context includes modifications to the type, extracted text, and relevancy to the inquiry for one or more factors in the results of the natural language processing of the context;

apply the modifications of the corrective information to adjust the results of the natural language processing of the context and modify the interpretation for the context; and

process the inquiry utilizing natural language processing with the modified interpretation for the context to adjust the interpretation for the inquiry;

retrieve content based on the adjusted interpretation for the inquiry and the context to produce candidate results for the inquiry; and

iteratively receive supplemental information for the inquiry to generate updated candidate results.

2. The computer program product of claim 1, wherein the inquiry requests medical information.

3. The computer program product of claim 1, wherein the context includes a structure with a schema defined for one of a problem class and a decision domain, wherein natural language utterances are processed to create the context aligned with the schema.

4. The computer program product of claim 1, wherein the processor presents the interpretation for the inquiry by:

presenting actions to be performed based on the interpretation for the inquiry to produce the candidate results and receiving modifications for the actions.

5. The computer program product of claim 1, wherein the supplemental information includes one or more from a group of:

information pertaining to the context;

user input adding, removing, or modifying the context to re-generate candidate results for a previous inquiry in light of altered context;

hypotheses proposed by the user to focus the candidate results; and

follow-up questions triggered from viewing the candidate results.

6. The computer program product of claim 1, wherein the computer readable program code further causes the processor to:

adapt processing to produce the candidate results based on processing of one or more prior inquiries.

7. The computer program product of claim 6, wherein the processor adapts processing to produce the candidate results by one or more from a group of:

collecting training information from the processing of the one or more prior inquiries and re-training models used to produce the candidate results;

collecting training information from processing each inquiry and dynamically adapting the models used to produce the candidate results;

adapting the processing based on user-provided corrections for the interpretation for one or more from a group of the context and the inquiry;

adapting the processing based on corrective inquiries from the user after receiving inferior candidate results;

receiving a rule set to adjust the processing; and

identifying common attributes of corrective information for the inquiry among users and generating models specific to one of a group and domain based on the common attributes to produce the candidate results.

8. The computer program product of claim 1, wherein the computer readable program code further causes the processor to:

determine a specific aspect of the context pertinent to the inquiry to retrieve the content.

9. A method of processing an inquiry to provide results satisfying the inquiry comprising:

processing a context using natural language processing to determine an interpretation for the context, wherein the context includes a description of a scenario pertaining to the inquiry and the interpretation for the context includes results of the natural language processing from one or more of parsing, entity detection, and relation detection of the context, and wherein the results of the natural language processing of the context include one or more factors each comprising a context portion including text extracted from the description of the scenario, and a type determined by the natural language processing for each factor comprising a description of the interpretation for the context portion corresponding to the factor;

processing the inquiry using natural language processing to determine an interpretation for the inquiry based on the interpretation for the context;

presenting the interpretation for the inquiry to a user;

in response to an incorrect interpretation for the inquiry;

receiving corrective information for the interpretation for the context from the user, wherein the corrective information for the interpretation for the context includes modifications to the type, extracted text, and relevancy to the inquiry for one or more factors in the results of the natural language processing of the context;

applying the modifications of the corrective information to adjust the results of the natural language processing of the context and modify the interpretation for the context; and

processing the inquiry utilizing natural language processing with the modified interpretation for the context to adjust the interpretation for the inquiry;

retrieving content based on the adjusted interpretation for the inquiry and the context to produce candidate results for the inquiry; and

iteratively receiving supplemental information for the inquiry to generate updated candidate results.

10. The method of claim 9, wherein the inquiry requests medical information.

11. The method of claim 9, wherein the context includes a structure with a schema defined for one of a problem class and a decision domain, wherein natural language utterances are processed to create the context aligned with the schema.

12. The method of claim 9, wherein presenting the interpretation for the inquiry further comprises:

presenting actions to be performed based on the interpretation for the inquiry to produce the candidate results and receiving modifications for the actions.

13. The method of claim 9, wherein the supplemental information includes one or more from a group of:

information pertaining to the context;

user input adding, removing, or modifying the context to re-generate candidate results for a previous inquiry in light of altered context;

hypotheses proposed by the user to focus the candidate results; and

follow-up questions triggered from viewing the candidate results.

14. The method of claim 9, further comprising:

adapting processing to produce the candidate results based on processing of one or more prior inquiries.

15. The method of claim 14, wherein adapting processing to produce the candidate results comprises one or more from a group of:

collecting training information from the processing of the one or more prior inquiries and re-training models used to produce the candidate results;

collecting training information from processing each inquiry and dynamically adapting the models used to produce the candidate results;

adapting the processing based on user-provided corrections for the interpretation of one or more from a group of the context and the inquiry;

adapting the processing based on corrective inquiries from the user after receiving inferior candidate results;

receiving a rule set to adjust the processing; and

identifying common attributes of corrective information for the inquiry among users and generating models specific to one of a group and domain based on the common attributes to produce the candidate results.

16. The method of claim 9, further comprising:

determining a specific aspect of the context pertinent to the inquiry to retrieve the content.

17. A system for processing an inquiry to provide results satisfying the inquiry comprising:

a computer system including at least one processor configured to;

process a context using natural language processing to determine an interpretation for the context, wherein the context includes a description of a scenario pertaining to the inquiry and the interpretation for the context includes results of the natural language processing from one or more of parsing, entity detection, and relation detection of the context, and wherein the results of the natural language processing of the context include one or more factors each comprising a context portion including text extracted from the description of the scenario, and a type determined by the natural language processing for each factor comprising a description of the interpretation for the context portion corresponding to the factor;

process the inquiry using natural language processing to determine an interpretation for the inquiry based on the interpretation for the context;

present the interpretation for the inquiry to a user;

in response to an incorrect interpretation for the inquiry;

receive corrective information for the interpretation for the context from the user, wherein the corrective information for the interpretation for the context includes modifications to the type, extracted text, and relevancy to the inquiry for one or more factors in the results of the natural language processing of the context;

apply the modifications of the corrective information to adjust the results of the natural language processing of the context and modify the interpretation for the context; and

process the inquiry utilizing natural language processing with the modified interpretation for the context to adjust the interpretation for the inquiry;

retrieve content based on the adjusted interpretation for the inquiry and the context to produce candidate results for the inquiry; and

iteratively receive supplemental information for the inquiry to generate updated candidate results.

18. The system of claim 17, wherein the at least one processor is configured to adapt processing to produce the candidate results based on processing of one or more prior inquiries by one or more from a group of:

collecting training information from the processing of the one or more prior inquiries and re-training models used to produce the candidate results;

collecting training information from processing each inquiry and dynamically adapting the models used to produce the candidate results;

adapting the processing based on user-provided corrections for the interpretation for one or more from a group of the context and the inquiry;

adapting the processing based on corrective inquiries from the user after receiving inferior candidate results;

receiving a rule set to adjust the processing; and

identifying common attributes of corrective information for the inquiry among users and generating models specific to one of a group and domain based on the common attributes to produce the candidate results.

19. The system of claim 17, wherein the context includes a structure with a schema defined for one of a problem class and a decision domain, wherein natural language utterances are processed to create the context aligned with the schema.

20. The system of claim 17, wherein the at least one processor is configured to present the interpretation for the inquiry by:

presenting actions to be performed based on the interpretation for the inquiry to produce the candidate results and receiving modifications for the actions.