Natural language understanding cache

• US 9,898,455 B2
• Filed: 12/01/2014
• Issued: 02/20/2018
• Est. Priority Date: 12/01/2014
• Status: Expired due to Fees

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Hebert, Matthieu, Gillet, Andre
• Primary Examiner(s): Ly, Anh
• Application Number:
  US14/556,904
• Publication Number:
  US 20160154783A1
• Time in Patent Office: 1,177 Days
• CPC Class Codes: G06F 12/0875 with dedicated cache, e.g. instruction or stack
• CPC Class Codes: G06F 16/907 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 16/9537 Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
• CPC Class Codes: G06F 2212/452 Instruction code
• CPC Class Codes: G06F 40/205 Parsing

Abstract

Disclosed methods and systems are directed to natural language understanding cache usage. The methods and systems may include receiving a first natural language input comprising a set of one or more terms, and parsing the first natural language input to determine a first pretag result comprising a first string comprising at least one term from the set of one or more terms. The methods and systems may also determine that if the first pretag result corresponds to a key stored in a cache, then retrieve one or more cached NLU results corresponding to the at least one key. The methods and systems may also determine that if the first pretag result does not correspond to a key stored in the cache, then determine, based on the set of one or more terms, a first NLU result corresponding to the first natural language input.

Claims

1. A method for performing natural language understanding (NLU) via computer processing, the method comprising:

determining, by a computing device, a preceding pretag result and a tentative NLU result, for each of one or more preceding natural language inputs;

storing, in a cache, one or more keys, each key comprising one preceding pretag result and a corresponding tentative NLU result;

receiving, by the computing device, a first natural language input comprising a set of one or more terms;

parsing, by the computing device, the first natural language input to determine a first pretag result comprising at least a first string comprising at least one term from the set of one or more terms, wherein the parsing comprises determining whether any of the one or more terms match metadata associated with a contacts list;

determining, by the computing device, whether the first pretag result corresponds to at least one key stored in the cache;

if the first pretag result corresponds to at least one key stored in the cache, retrieving, by the computing device and from the cache, one or more cached NLU results corresponding to the at least one key; and

if the first pretag result does not correspond to at least one key stored in the cache;

determining, by the computing device and based on the set of one or more terms, a first NLU result corresponding to the first natural language input, the first NLU result comprising an intent associated with the first natural language input;

storing, by the computing device and in the cache, the first NLU result; and

storing, by the computing device and in the cache, a first key comprising the first pretag result, the first key corresponding to the first NLU result.

2. The method of claim 1, wherein the first pretag result corresponds to at least one key stored in the cache, the method further comprising determining, by the computing device and for the first natural language input, a final NLU result based on the first pretag result and the one or more cached NLU results.

3. The method of claim 1, wherein the first pretag result does not correspond to at least one key stored in the cache, and wherein the cache stores the first NLU result and the first key comprising the first pretag result, the method further comprising:

receiving, by the computing device, a second natural language input comprising one or more terms;

parsing, by the computing device, the second natural language input to determine a second pretag result;

determining, by the computing device, that the second pretag result corresponds to the first key stored in the cache; and

retrieving, by the computing device and from the cache the first NLU result.

4. The method of claim 1, wherein parsing the first natural language input to determine the first pretag result comprises determining, by the computing device, a named entity based on the one or more terms comprised in the first natural language input, the named entity comprising at least one of the one or more terms comprised in the first natural language input.

5. The method of claim 4, wherein the first string comprising at least one term of the set of one or more terms further comprises the metadata.

6. The method of claim 1, wherein the at least one key comprises one or more of the following:

a named entity identified from the one or more terms comprised in the first natural language input; and

a string of words comprising at least one of the one or more terms comprised in the first natural language input.

7. The method of claim 1, wherein the determining the preceding pretag result for the one or more preceding natural language inputs comprises determining, based on device or user metadata, at least one string of one or more terms and wherein the determining the tentative NLU result for the one or more preceding natural language inputs comprises determining an intent associated with the one or more preceding natural language inputs, a string of terms, or a named entity.

8. The method of claim 1, wherein parsing the first natural language input to determine the first pretag result further comprises parsing, by the computing device, the first natural language input based on additional metadata to determine the first pretag result, and wherein the additional metadata comprises metadata associated with at least one of:

a list of favorite websites, a favorite telephone number list, a favorite mobile applications list, a favorite restaurants list, a favorite business list, or a favorite location list.

9. A system, comprising:

at least one processor; and

at least one memory storing instructions that, when executed by the at least one processor, cause the system to perform;

determining, by a device, a preceding pretag result and a tentative NLU result, for each of one or more preceding natural language inputs;

storing, in a cache, one or more keys, each key comprising one preceding pretag result and a corresponding tentative NLU result;

receiving, by the device, a first natural language input comprising a set of one or more terms;

parsing the first natural language input to determine a first pretag result comprising at least a first string comprising at least one term from the set of one or more terms, wherein the parsing comprises determining whether any of the one or more terms match metadata associated with a contacts list;

determining whether the first pretag result corresponds to at least one key stored in the cache;

if the first pretag result corresponds to at least one key stored in the cache, retrieving, from the cache, one or more cached NLU results corresponding to the at least one key; and

if the first pretag result does not correspond to at least one key stored in the cache;

determining, based on the set of one or more terms, a first NLU result corresponding to the first natural language input, the first NLU result comprising an intent associated with the first natural language input;

storing, in the cache, the first NLU result; and

storing, in the cache, a first key comprising the first pretag result, the first key corresponding to the first NLU result.

10. The system of claim 9, wherein the first pretag result corresponds to at least one key stored in the cache, the instructions further cause the system to perform determining, for the first natural language input, a final NLU result based on the first pretag result and the one or more cached NLU results.

11. The system of claim 9, wherein the first pretag result does not correspond to at least one key stored in the cache, and wherein the cache stores the first NLU result and the first key comprising the first pretag result, the instructions further cause the system to perform:

receiving a second natural language input comprising one or more terms;

parsing the second natural language input to determine a second pretag result;

determining that the second pretag result corresponds to the first key stored in the cache; and

retrieving from the cache the first NLU result.

12. The system of claim 9, wherein parsing the first natural language input to determine the first pretag result comprises determining a named entity based on the one or more terms comprised in the first natural language input, the named entity comprising at least one of the one or more terms comprised in the first natural language input.

13. The system of claim 9, wherein the at least one key comprises one or more of the following:

a named entity identified from the one or more terms comprised in the first natural language input; and

a string of words comprising at least one of the one or more terms comprised in the first natural language input.

14. The system of claim 9, wherein the instructions further cause the system to perform:

determining the preceding pretag result for the one or more preceding natural language inputs by determining, based on device or user associated metadata, at least one string of one or more terms; and

determining the tentative NLU result for the one or more preceding natural language inputs by determining an intent associated with the one or more preceding natural language inputs, a string of terms, or a named entity.

15. The system of claim 9, wherein parsing the first natural language input to determine the first pretag result further comprises parsing the first natural language input based on additional metadata to determine the first pretag result, and wherein the additional metadata comprises metadata associated with at least one of:

a list of favorite websites, a favorite telephone number list, a favorite mobile applications list, a favorite restaurants list, a favorite business list, or a favorite location list.

16. One or more non-transitory computer-readable storage media having instructions stored thereon, that when executed by one or more processors, cause the one or more processors to perform:

determining, by a device, a preceding pretag result and a tentative NLU result, for each of one or more preceding natural language inputs;

storing, in a cache, one or more keys, each key comprising one preceding pretag result and a corresponding tentative NLU result;

receiving, by the device, a first natural language input comprising a set of one or more terms;

parsing the first natural language input to determine a first pretag result comprising at least a first string comprising at least one term from the set of one or more terms, wherein the parsing comprises determining whether any of the one or more terms match metadata associated with a contacts list;

determining whether the first pretag result corresponds to at least one key stored in the cache;

if the first pretag result corresponds to at least one key stored in the cache, retrieving, from the cache, one or more cached NLU results corresponding to the at least one key; and

if the first pretag result does not correspond to at least one key stored in the cache;

determining, based on the set of one or more terms, a first NLU result corresponding to the first natural language input, the first NLU result comprising an intent associated with the first natural language input;

storing, in the cache, the first NLU result; and

storing, in the cache, a first key comprising the first pretag result, the first key corresponding to the first NLU result.

17. The one or more non-transitory computer-readable storage media of claim 16, wherein the first pretag result corresponds to at least one key stored in the cache, the instructions further cause the one or more processors to perform determining, for the first natural language input, a final NLU result based on the first pretag result and the one or more cached NLU results.

18. The one or more non-transitory computer-readable storage media of claim 16, wherein the first pretag result does not correspond to at least one key stored in the cache, and wherein the cache stores the first NLU result and the first key comprising the first pretag result, the instructions further cause the one or more processors to perform:

receiving a second natural language input comprising one or more terms;

parsing the second natural language input to determine a second pretag result;

determining that the second pretag result corresponds to the first key stored in the cache; and

retrieving from the cache the first NLU result.

19. The one or more non-transitory computer-readable storage media of claim 16, wherein parsing the first natural language input to determine the first pretag result comprises determining a named entity based on the one or more terms comprised in the first natural language input, the named entity comprising at least one of the one or more terms comprised in the first natural language input.

20. The one or more non-transitory computer-readable storage media of claim 16, wherein parsing the first natural language input to determine the first pretag result further comprises parsing the first natural language input based on additional metadata to determine the first pretag result, and wherein the additional metadata comprises metadata associated with at least one of:

a list of favorite websites, a favorite telephone number list, a favorite mobile applications list, a favorite restaurants list, a favorite business list, or a favorite location list.