In-context access of stored declarative knowledge using natural language expression

• US 9,805,020 B2
• Filed: 02/06/2015
• Issued: 10/31/2017
• Est. Priority Date: 04/23/2009
• Status: Active Grant

• Current Assignee: Gorman Family Trust
• Original Assignee: Deep Sky Concepts, Inc.
• Inventors: Gorman, John G., Gorman, John W. R.
• Primary Examiner(s): Perveen, Rehana
• Assistant Examiner(s): Tran, Loc
• Application Number:
  US14/616,607
• Publication Number:
  US 20150186504A1
• Time in Patent Office: 998 Days
• Field of Search: 707771, 707769, 707767
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Identifying a context for parsing a natural language expression. In an embodiment, a table comprising rows addressed according to context compartments is disclosed. Each compartment represents a context and comprises row(s), and each row represents a concept and comprises an outline field indicating an order of the concept in its context. Input expression(s) are received, and a first search is performed on the table to identify rows representing concepts corresponding to element(s) of the expression(s) and rows representing concepts corresponding to combinations of concepts represented by previously identified rows until a first row representing a higher-order concept is identified. Based on the first row, a first context compartment is determined, and a second search, that is restricted to the first context compartment, is performed to identify a second row representing a concept corresponding to an entirety of at least one of the input expression(s).

Claims

1. A method for identifying a context for parsing a natural language expression, the method comprising using at least one hardware processor to:

interface with a database comprising at least one table, wherein the at least one table comprises a plurality of rows addressed according to a plurality of context compartments, wherein each of the plurality of context compartments represents a context and comprises a subset of one or more of the plurality of rows, and wherein each of the plurality of rows represents a concept and comprises an outline field that comprises an indication of an order of that concept in the context represented by the context compartment that comprises the row;

receive one or more input expressions comprising a plurality of language elements, wherein the one or more input expressions comprise a first input expression and a second input expression;

perform a first search that comprises searching the at least one table to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a first row representing a higher-order concept corresponding to an entirety of the first input expression is identified;

based on the identified first row, set a first one of the plurality of context compartments as a current context compartment for subsequent searching; and

,

after setting the first context compartment as the current context compartment, perform a restricted second search that comprises only searching the first context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a second row representing a concept corresponding to an entirety of the second input expression is identified or no such second row is determined to exist in the first context compartment,

wherein each of the plurality of context compartments is associated with at least one of a plurality of restricted dictionaries,

wherein each of the plurality of restricted dictionaries comprises a subset of words from a full dictionary,

wherein the second input expression is an audio input that is received via a speech-to-text engine after the first context compartment has been set as the current context compartment, and

wherein the speech-to-text engine uses only the subset of words from one of the plurality of restricted dictionaries that is associated with the current context compartment to convert the second input expression into text.

2. The method of claim 1, wherein setting the first context compartment as the current context compartment comprises identifying the first context compartment as a context compartment that comprises the identified first row.

3. The method of claim 2, the identification of the first context compartment as a context compartment that comprises the identified first row is based on a value in the outline field of the identified first row, and wherein only searching the current context compartment comprises restricting the search to only those of the plurality of rows in the at least one table that comprise outline fields having the same value as the value in the outline field of the identified first row.

4. The method of claim 2, wherein identifying the first context compartment comprises identifying a range of rows that include the identified first row.

5. The method of claim 1, wherein the identified first row comprises a context identifier field which comprises an identifier of the first context compartment, and wherein the first context compartment is different than a context compartment which comprises the identified first row.

6. The method of claim 1, further comprising using the at least one hardware processor to, when no second row is determined to exist in the first context compartment during the restricted second search, perform a relaxed third search that comprises searching at least a second one of the plurality of context compartments to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a second row representing a concept corresponding to the entirety of the second input expression is identified or no such second row is determined to exist in the second context compartment.

7. The method of claim 6, wherein the first context compartment is nested within the second context compartment.

8. The method of claim 7, wherein the second context compartment comprises all of the plurality of rows in the at least one table.

9. The method of claim 6, wherein the first context compartment and the second context compartment do not overlap.

10. The method of claim 1, further comprising using the at least one hardware processor to sort the plurality of rows in the at least one table by the outline field to arrange each of the sorted plurality of rows such that it is grouped with other ones of the plurality of rows in its same context compartment.

11. The method of claim 10, wherein, for each of the plurality of rows in the at least one table, the outline field in the row comprises a value which identifies both a macro context compartment that comprises the row and a micro context compartment that comprises the row, wherein the macro context compartment comprises the same subset of one or more rows that the micro context compartment comprises and one or more additional ones of the plurality of rows.

12. The method of claim 11, wherein the value in each outline field comprises a first number that identifies the macro context compartment and a second number that identifies the micro context compartment.

13. The method of claim 1, further comprising using the at least one hardware processor to, when a plurality of rows corresponding to the higher-order concept are identified during the first search and the plurality of rows are in separate ones of the plurality of context compartments, prompt a user for contextual information.

14. The method of claim 1, further comprising using the at least one hardware processor to maintain a record of context compartments searched during the first search and the second search.

15. The method of claim 1, wherein the first input expression is a heading, wherein the first context compartment comprises a subset of two or more rows representing concepts associated with the heading, and wherein the method comprises performing the first search in a heading context compartment, which comprises a subset of two or more rows representing a plurality of heading concepts, and wherein the identified first row comprises a context pointer field that identifies the first context compartment.

16. The method of claim 1, wherein the second row comprises a field indicating a first response to the at least one input expression, and wherein the method further comprises, once the second row is identified, generating the first response.

17. The method of claim 16, wherein the concept corresponding to an entirety of the at least one input expression is represented by both the second row in the first context compartment and a third row in a second one of the plurality of context compartments, wherein the third row comprises a field indicating a second response to the at least one input expression, and wherein the second response is different than the first response.

18. The method of claim 17, further comprising using the at least one hardware processor to, at a time that is different than a time during which the first search and the second search were performed:

receive the at least one input expression again, wherein the at least one input expression comprises a plurality of language elements;

perform a third search that comprises searching the second context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements of the subsequent input expression and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until the third row representing the concept corresponding to the entirety of the at least one input expression is identified; and

generate the second response which is different than the first response.

19. The method of claim 1, wherein the concept corresponding to the entirety of the at least one input expression can be expressed as any of a plurality of distinct expressions including the at least one input expression, wherein the first context compartment comprises a plurality of rows linked together as a plurality of binary trees with the second row as a root of each of the plurality of binary trees, and wherein each of the plurality of binary trees represents one of the plurality of distinct expressions, such that performing the second search for each of the plurality of distinct expressions will result in identification of the second row.

20. The method of claim 1, further comprising using the at least one hardware processor to, when the second row is identified during the second search in the first context compartment:

receive a subsequent input expression comprising a plurality of language elements; and

perform a restricted third search that comprises only searching the first context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements of the subsequent input expression and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a third row representing a concept corresponding to an entirety of the subsequent input expression is identified or no such third row is determined to exist in the first context compartment.

21. The method of claim 1, wherein each of the plurality of rows in the at least one table comprise a plurality of outline fields, wherein each of the plurality of outline fields comprises an indication of an order of the concept represented by the row according to an ordering basis that is different than any of the other plurality of outline fields, and wherein the method further comprises:

performing a first ordering of the plurality of rows according to a first subset of one or more of the plurality of outline fields; and

performing a second ordering of the plurality of rows according to a second subset of one or more of the plurality of outline fields, wherein the second subset is different than the first subset.

22. The method of claim 1, wherein the plurality of context compartments comprise a screen context compartment, and wherein the method further comprises using the at least one hardware processor to, prior to the first search:

perform a screening search that comprises searching the screener context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a screening row representing a concept corresponding to an entirety of at least one of the one or more input expressions is identified or no such screening row is determined to exist in the screener context compartment;

when the screening row is identified, prompt a user for contextual information; and

,

when no such screening row is determined to exist in the screener context compartment, perform the first search.

23. The method of claim 1, wherein the plurality of context compartments comprise a common sense context compartment which comprises a subset of two or more rows that represent one or more common concepts that are common to two or more of the plurality of context compartments, and wherein the method further comprises using the at least one hardware processor to, when no second row is determined to exist in the first context compartment during the restricted second search, perform a third search that comprises searching the common sense context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements or to combinations of two or more concepts represented by previously identified rows until a third row representing a concept that is combinable with another concept of a previously identified row to form a higher-order concept represented by a fourth row in the first context compartment is identified or no such third row is determined to exist in the common sense context compartment.

24. A system for identifying a context for parsing a natural language expression, the system comprising:

at least one hardware processor;

a database comprising at least one table, wherein the at least one table comprises a plurality of rows addressed according to a plurality of context compartments, wherein each of the plurality of context compartments represents a context and comprises a subset of one or more of the plurality of rows, and wherein each of the plurality of rows represents a concept and comprises an outline field that comprises an indication of an order of that concept in the context represented by the context compartment that comprises the row; and

one or more modules configured to, when executed by the at least one hardware processor,

receive one or more input expressions comprising a plurality of language elements, wherein the one or more input expressions comprise a first input expression and a second input expression,

perform a first search that comprises searching the at least one table to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a first row representing a higher-order concept corresponding to an entirety of the first input expression is identified,

based on the identified first row, set a first one of the plurality of context compartments as a current context compartment for subsequent searching, and,

after setting the first context compartment as the current context compartment, perform a restricted second search that comprises only searching the first context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a second row representing a concept corresponding to an entirety of the second input expression is identified or no such second row is determined to exist in the first context compartment,

wherein each of the plurality of context compartments is associated with at least one of a plurality of restricted dictionaries,

wherein each of the plurality of restricted dictionaries comprises a subset of words from a full dictionary,

wherein the second input expression is an audio input that is received via a speech-to-text engine after the first context compartment has been set as the current context compartment, and

wherein the speech-to-text engine uses only the subset of words from one of the plurality of restricted dictionaries that is associated with the current context compartment to convert the second input expression into text.

25. A non-transitory computer-readable medium having instructions stored thereon, wherein the instructions, when executed by a processor, cause the processor to:

interface with a database comprising at least one table, wherein the at least one table comprises a plurality of rows addressed according to a plurality of context compartments, wherein each of the plurality of context compartments represents a context and comprises a subset of one or more of the plurality of rows, and wherein each of the plurality of rows represents a concept and comprises an outline field that comprises an indication of an order of that concept in the context represented by the context compartment that comprises the row;

receive one or more input expressions comprising a plurality of language elements, wherein the one or more input expressions comprise a first input expression and a second input expression;

perform a first search that comprises searching the at least one table to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a first row representing a higher-order concept corresponding to an entirety of the first input expression is identified;

based on the identified first row, set a first one of the plurality of context compartments as a current context compartment for subsequent searching; and

,

after setting the first context compartment as the current context compartment, perform a restricted second search that comprises only searching the first context compartment to identify rows representing concepts that correspond to one or more of the plurality of language elements and rows representing concepts that correspond to combinations of two or more concepts represented by previously identified rows until a second row representing a concept corresponding to an entirety of the second input expression is identified or no such second row is determined to exist in the first context compartment,

wherein each of the plurality of context compartments is associated with at least one of a plurality of restricted dictionaries,

wherein each of the plurality of restricted dictionaries comprises a subset of words from a full dictionary,

wherein the second input expression is an audio input that is received via a speech-to-text engine after the first context compartment has been set as the current context compartment, and

wherein the speech-to-text engine uses only the subset of words from one of the plurality of restricted dictionaries that is associated with the current context compartment to convert the second input expression into text.