CONTEXTUAL SPOKEN LANGUAGE UNDERSTANDING IN A SPOKEN DIALOGUE SYSTEM

• US 20190066668A1
• Filed: 08/25/2017
• Published: 02/28/2019
• Est. Priority Date: 08/25/2017
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: LIN, Xihui, McNAMARA, Andrew James, HE, Jing
• Granted Patent: US 11,081,106 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/295 Named entity recognition
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 5/027 Frames
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/14 using statistical models, e.g. Hidden Markov Models [HMMs] G10L15/18 takes precedence
• CPC Class Codes: G10L 15/16 using artificial neural networks
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

A spoken dialogue system includes a spoken language understanding apparatus. The spoken language understanding apparatus can include an intent apparatus and a selection apparatus. The intent apparatus is configured to determine if a query comprises a global command, to determine if an intent associated with a query is or is not included in a domain that is supported by the spoken dialogue system, to determine if a query comprises a confirmation type, to tag one or more entities in a query, and to determine an intent probability distribution and a domain probability distribution that is associated with a query. When the query includes an entity that is included in two or more possible entities, the selection apparatus is configured to provide a score for each of the two or more possible entities.

Claims

1. A spoken dialogue system, comprising:

a global apparatus for determining the query comprises a global command;

an out-of-domain apparatus for determining if the query is an out-of-domain query that is not supported by the spoken dialogue system;

a confirmation apparatus for determining if the query comprises a confirmation type; and

an intent and domain apparatus for determining an intent probability distribution and a domain probability distribution for the query.

2. The spoken dialogue system of claim 1, further comprising an entity apparatus for tagging each entity in the query.

3. The spoken dialogue system of claim 2, wherein the global apparatus, the out-of-domain apparatus, the confirmation apparatus, the intent and domain apparatus, and the entity apparatus are included in an intent apparatus in the system.

4. The spoken dialogue system of claim 3, further comprising a selection apparatus operably connected to the intent apparatus, wherein the selection apparatus provides a score for each candidate option when a tagged entity is associated with multiple candidate options.

5. The spoken dialogue system of claim 4, wherein the multiple candidate options comprise multiple candidate contacts.

6. The spoken dialogue system of claim 4, wherein the multiple candidate options comprise multiple candidate places.

7. The spoken dialogue system of claim 1, wherein the confirmation type is one of multiple confirmation types.

8. The spoken dialogue system of claim 1, wherein the global command is one of multiple global commands.

9. A system for operating a spoken dialogue system, the system comprising:

a processing unit; and

a storage device for storing computer executable instructions that when performed by the processing unit, perform a method comprising;

receiving, by the spoken dialogue system, a query;

processing, by a spoken language apparatus in the spoken dialogue system, the query to;

determine if the query is an out-of-domain query that is not supported by the spoken dialogue system;

determine if the query comprises a global command; and

determine if the query comprises a confirmation type.

10. The system of claim 9, wherein the query is processed by the spoken language understanding apparatus to tag a previous intent and each entity in the query.

11. The system of claim 10, wherein the query is processed by the spoken language understanding apparatus to determine a probability distribution over all intents and a probability distribution over all domains based at least in part on each tagged entity and the tagged previous intent.

12. The system of claim 9, further comprising performing a selection operation when the query includes an entity that is one of two or more possible entities, wherein the selection operation produces a score for each possible entity that is included in the two or more possible entities.

13. The system of claim 12, wherein the two or more possible entities comprise two or more contacts.

14. The system of claim 12, wherein the two or more possible entities comprise two or more places.

15. The system of claim 12, further comprising prior to performing the selection operation, providing the two or more possible entities to an output device and responsively receiving a second query.

16. The system of claim 9, wherein the confirmation type is one of multiple confirmation types.

17. The system of claim 9, wherein the global command is one of multiple global commands.

18. The system of claim 9, wherein the system comprises a server-computing device and the query is received from a client-computing device.

19. A method of providing a spoken language understanding apparatus that includes a global apparatus for determining the query comprises a global command, an entity apparatus for tagging one or more entities in the query, an intent and domain apparatus for determining an intent and a domain associated with the query, and a confirmation apparatus for determining the query comprises a confirmation type, the method comprising:

determining initial word and phrase embeddings using a skip-gram training process;

adjusting the initial word and phrase embeddings using a plurality of lists to produce a first set of adjusted embeddings, each list comprising words and phrases that share similar semantic representations;

training the spoken language understanding apparatus using the first set of adjusted embeddings; and

during training, adjusting the first set of adjusted embeddings to produce a second set of adjusted embeddings, wherein the first set of adjusted embeddings are adjusted to improve a correctness of an output produced by at least one of the global apparatus, the entity apparatus, the intent and domain apparatus, or the confirmation apparatus.

20. The method of claim 19, wherein the global apparatus, the entity apparatus, the intent and domain apparatus, and the confirmation apparatus are included in a neural network and the method further comprises adjusting one or more parameters of the neural network to improve a correctness of an output produced by at least one of the global apparatus, the entity apparatus, the intent and domain apparatus, or the confirmation apparatus.