Enhanced knowledge repository

• US 8,838,659 B2
• Filed: 09/29/2008
• Issued: 09/16/2014
• Est. Priority Date: 10/04/2007
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Tunstall-Pedoe, William
• Primary Examiner(s): BURKE, JEFF A
• Application Number:
  US12/240,699
• Publication Number:
  US 20090192968A1
• Time in Patent Office: 2,178 Days
• Field of Search: 706/12, 706 45- 46, 706/55, 706/59, 706/61
• US Class Current: 707/899
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

Embodiments of the present invention relate to knowledge representation systems which include a knowledge base in which knowledge is represented in a structured, machine-readable format that encodes meaning.

Claims

1. A knowledge representation system, comprising:

a data store having a knowledge base stored therein comprising knowledge represented in a structured, machine-readable format which encodes meaning and is distinct from natural language, the structured, machine-readable format including entity objects, each of at least some of the entity objects being explicitly associated in the knowledge base with at least one of a plurality of knowledge domains; and

at least one computing device configured to;

translate a natural language question to a plurality of queries, the natural language question being associated with a first knowledge domain of the plurality of knowledge domains, each of the plurality of queries being a machine-readable representation of a possible translation of the natural language question that is compatible with the structured, machine-readable format of the knowledge base;

determine whether any of the queries relate to the first knowledge domain by determining if there are any named entities in the queries that correspond to entity objects that are associated with the first knowledge domain in the knowledge base, thereby identifying a first subset of the queries;

determine whether any of the first subset of the queries represents a corresponding interpretation of the natural language question that belongs to the first knowledge domain, thereby identifying a second subset of the queries; and

respond to the natural language question using a portion of the knowledge within the first knowledge domain by processing the second subset of the queries that relate to the first knowledge domain with reference to the knowledge base.

2. The system of claim 1 wherein the at least one computing device is further configured to facilitate disambiguation of the plurality of queries.

3. The system of claim 1 wherein the at least one computing device is further configured to generate a plurality of answers corresponding to the plurality of queries.

4. The system of claim 1 wherein the at least one computing device is further configured to exclude another portion of the knowledge not within the first knowledge domain from generation of one or more answers to the second subset of the queries.

5. The system of claim 1 wherein the first knowledge domain is defined by an end user, and wherein the at least one computing device is further configured to create a sub-domain of the knowledge base corresponding to the first knowledge domain, and to respond to all subsequent queries corresponding to the sub-domain using the portion of the knowledge within the first knowledge domain.

6. A knowledge representation system, comprising:

a data store having a knowledge base stored therein comprising knowledge represented in a structured, machine-readable format which encodes meaning and is distinct from natural language, the knowledge base including a plurality of existing facts each of which is represented in the machine-readable format, and a plurality of generators configured to generate new facts not present in the knowledge base, the knowledge base also including a plurality of explanations each of which is identified as corresponding to one or more of the facts or one or more of the generators, the explanations including human-readable explanatory text relating to the corresponding facts or generators; and

at least one computing device configured to;

identify one or more of the existing facts from the knowledge base corresponding to a query;

identify one or more of the generators from the knowledge base corresponding to the query;

generate one or more new facts using the one or more generators;

identify one or more of the explanations from the knowledge base corresponding to the one or more existing facts;

identify one or more of the explanations from the knowledge base corresponding to the one or more generators by which the one or more new facts were generated;

construct a linked list of the explanations corresponding to the one or more existing facts and the one or more generators by which the one or more new facts were generated;

remove a redundant explanation from the linked list of explanations resulting in a modified linked list of explanations;

respond to the query using the explanatory text corresponding to the modified linked list of explanations to explain how the response was derived.

7. The system of claim 6 wherein the at least one computing device is further configured to translate a natural language question to the query, the natural language question comprising one of a how question, a why question, or a request for a definition of a term.

8. The system of claim 6 wherein some of the explanations comprise language translations of others of the explanations, each of the explanations having one or more corresponding translations identifying the others of the explanations to which it corresponds.

9. The system of claim 6 wherein the at least one computing device is further configured to:

generate a modified version of a particular one of the explanations;

associate the modified version of the particular explanation with the particular explanation and infer that the one or more facts corresponding to the particular explanation also correspond to the modified version of the particular explanation where the modified version of the particular explanation represents a first class of modification; and

associate the one or more facts corresponding to the particular explanation with the modified version of the particular explanation where the modified version of the particular explanation represents a second class of modification.

10. The system of claim 9 where the at least one computing device is further operable to generate the modified version of the particular explanation in response to modifications entered by a user, the modifications including modification class information entered by the user indicating whether the modified version of the particular explanation represents the first or second class of modification.

11. A computer-implemented method for use with a knowledge representation system, the knowledge representation system including a data store having a knowledge base stored therein comprising knowledge represented in a structured, machine-readable format which encodes meaning and is distinct from natural language, the structured, machine-readable format including entity objects, each of at least some of the entity objects being explicitly associated in the knowledge base with at least one of a plurality of knowledge domains, the method comprising:

translating a natural language question to a plurality of queries, the natural language question being associated with a first knowledge domain of the plurality of knowledge domains, each of the plurality of queries being a machine-readable representation of a possible translation of the natural language question that is compatible with the structured, machine-readable format of the knowledge base;

determining whether any of the queries relate to the first knowledge domain by determining if there are any named entities in the queries that correspond to entity objects that are associated with the first knowledge domain in the knowledge base, thereby identifying a first subset of the queries;

determining whether any of the first subset of the queries represents a corresponding interpretation of the natural language question that belongs to the first knowledge domain, thereby identifying a second subset of the queries; and

responding to the natural language question using a portion of the knowledge within the first knowledge domain by processing the second subset of the queries that relate to the first knowledge domain with reference to the knowledge base.

12. The method of claim 11 further comprising facilitating disambiguation of the plurality of queries.

13. The method of claim 11 further comprising generating a plurality of answers corresponding to the plurality of queries.

14. The method of claim 11 further comprising excluding another portion of the knowledge not within the first knowledge domain from generation of one or more answers to the second subset of the queries.

15. The method of claim 11 wherein the first knowledge domain is defined by an end user, and wherein the method further comprises creating a sub-domain of the knowledge base corresponding to the first knowledge domain, and responding to all subsequent queries corresponding to the sub-domain using the portion of the knowledge within the first knowledge domain.

16. A computer-implemented method for use with a knowledge representation system, the knowledge representation system including a data store having a knowledge base stored therein comprising knowledge represented in a structured, machine-readable format which encodes meaning and is distinct from natural language, the knowledge base including a plurality of existing facts each of which is represented in the machine-readable format, and a plurality of generators configured to generate new facts not present in the knowledge base, the knowledge base also including a plurality of explanations each of which is identified as corresponding to one or more of the facts or one or more of the generators, the explanations including human-readable explanatory text relating to the corresponding facts or generators, the method comprising:

identifying one or more of the existing facts from the knowledge base corresponding to a query;

identifying one or more of the generators from the knowledge base corresponding to the query;

generating one or more new facts using the one or more generators;

identifying one or more of the explanations from the knowledge base corresponding to the one or more existing facts;

identifying one or more of the explanations from the knowledge base corresponding to the one or more generators by which the one or more new facts were generated;

constructing a linked list of the explanations corresponding to the one or more existing facts and the one or more generators by which the one or more new facts were generated;

removing a redundant explanation from the linked list of explanations resulting in a modified linked list of explanations; and

responding to the query using the explanatory text corresponding to the modified linked list of explanations to explain how the response was derived.

17. The method of claim 16 further comprising translating a natural language question to the query, the natural language question comprising one of a how question, a why question, or a request for a definition of a term.

18. The method of claim 16 wherein some of the explanations comprise language translations of others of the explanations, each of the explanations having one or more corresponding translations identifying the others of the explanations to which it corresponds.

19. The method of claim 16 further comprising:

generating a modified version of a particular one of the explanations;

associating the modified version of the particular explanation with the particular explanation;

inferring that the one or more facts corresponding to the particular explanation also correspond to the modified version of the particular explanation where the modified version of the particular explanation represents a first class of modification; and

associating the one or more facts corresponding to the particular explanation with the modified version of the particular explanation where the modified version of the particular explanation represents a second class of modification.

20. The method of claim 19 further comprising generating the modified version of the particular explanation in response to modifications entered by a user, the modifications including modification class information entered by the user indicating whether the modified version of the particular explanation represents the first or second class of modification.