Customer service system, method, and software program product for responding to queries using natural language understanding

• US 8,923,502 B2
• Filed: 02/22/2013
• Issued: 12/30/2014
• Est. Priority Date: 06/24/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Sharp, Robert Douglas
• Primary Examiner(s): Nguyen, Duc
• Assistant Examiner(s): Mohammed, Assad
• Application Number:
  US13/774,910
• Publication Number:
  US 20130223600A1
• Time in Patent Office: 676 Days
• Field of Search: None
• US Class Current: 379/265.01
• CPC Class Codes: H04M 2201/40 using speech recognition
• CPC Class Codes: H04M 2203/558 Databases
• CPC Class Codes: H04M 3/00 Automatic or semi-automatic exchanges
• CPC Class Codes: H04M 3/5166 in combination with interactive voice response systems or voice portals, e.g. as front-ends
• CPC Class Codes: H04M 3/5231 with call back arrangements

Abstract

The present invention uses natural language understanding to increase the ability of a customer service system to respond to a user'"'"'s query in an automated manner. A customer service system receives a query from a user and offers the user the option of having the system contact the user at a later time with an answer. If the user accepts the offer, the customer service system processes the query offline, including providing the query to a natural language understanding interpreter. The system uses the natural language understanding interpretation to determine if the user'"'"'s query is in a database of frequently-asked queries. For each query in the database of frequently-asked queries, there is a predetermined response protocol. If the user'"'"'s query substantially matches a query in the database, the IVR system contacts the user with an automated response in accordance with the predetermined response protocol for the query.

Claims

1. A method, performed by an IVR system, for responding to users'"'"' queries while minimizing agent cost, the method comprising:

receiving a call from a user at an IVR system;

receiving a request from the user to speak with a live agent;

queuing the user for a live agent;

prompting the user to state a query;

recording the user'"'"'s speech input for the query;

providing the user'"'"'s query to a natural language understanding interpreter;

receiving a natural language interpretation result;

using the natural language interpretation result, determining if the user'"'"'s query matches a query in a database that includes frequently-asked queries and corresponding response protocols;

in response to the user'"'"'s query matching a query in the database, providing an automated response to the user in accordance with a predetermined response protocol for the query and asking the user if the user still desires to speak with a live agent;

in response to the user'"'"'s query not matching a query in the database or in response to the user indicating that the user still desires to speak with a live agent, transferring the user to a live agent;

in response to the user being transferred to a live agent and after the live agent provides the user with a response to the user'"'"'s query, prompting the user for feedback on the live agent'"'"'s response, grading the live agent'"'"'s response based on the user'"'"'s feedback, and storing the live agent'"'"'s response in association with the grade; and

selecting a new response protocol for the query based at least in part on the grades associated with live agents'"'"' responses to the query.

2. The method of claim 1, wherein:

the natural language interpretation result comprises semantic slots and corresponding values;

the database of frequently-asked queries includes semantic slots and corresponding values for frequently-asked queries; and

the user'"'"'s query is determined to be in the database if the semantic slots and corresponding values in the interpretation result substantially match an entry in the database with a minimum confidence threshold.

3. The method of claim 1, further comprising:

for users'"'"' queries not in the database, ranking each type of user query based on the amount of live agent time consumed in handling the query; and

using the rankings, prioritizing queries for adding to the database.

4. The method of claim 1, wherein if the user'"'"'s query does not match a query in the database or the user indicates that the user still desires to speak with a live agent, the user is given a high priority for a live agent.

5. The method of claim 1, wherein the automated response to the user in accordance with the predetermined response protocol is provided via a voice interface.

6. The method of claim 1, wherein the automated response to the user in accordance with the predetermined response protocol is provided via SMS.

7. The method of claim 1, wherein the automated response to the user in accordance with the predetermined response protocol is provided via email.

8. The method of claim 1, wherein the automated response to the user in accordance with the predetermined response protocol is provided via web chat.

9. The method of claim 1, wherein the response protocol identifies any information needed to respond to the query and, if the IVR system does not have all such information, the IVR system'"'"'s initial response to the user is to request more information from the user.

10. The method of claim 9, wherein, after receiving requested information from the user, the IVR system provides the user with a response to the user'"'"'s query in accordance with the response protocol for the query.

11. A computer program embodied on a non-transitory computer-readable storage medium and comprising code, that, when executed by a computer system, enables the computer system to perform the following method:

receiving a call from a user at an IVR system;

receiving a request from the user to speak with a live agent;

queuing the user for a live agent;

prompting the user to state a query;

recording the user'"'"'s speech input for the query;

providing the user'"'"'s query to a natural language understanding interpreter;

receiving a natural language interpretation result;

using the natural language interpretation result, determining if the user'"'"'s query matches a query in a database that includes frequently-asked queries and corresponding response protocols;

in response the user'"'"'s query matching a query in the database, providing an automated response to the user in accordance with a predetermined response protocol for the query and asking the user if the user still desires to speak with a live agent;

in response to the user'"'"'s query not matching a query in the database or in response to the user indicating that the user still desires to speak with a live agent, transferring the user to a live agent;

in response to the user being transferred to a live agent and after the live agent provides the user with a response to the user'"'"'s query, prompting the user for feedback on the live agent'"'"'s response, grading the live agent'"'"'s response based on the user'"'"'s feedback, and storing the live agent'"'"'s response in association with the grade; and

selecting a new response protocol for the query based at least in part on the grades associated with live agents'"'"' responses to the query.

12. The computer program of claim 11, wherein:

the natural language interpretation result comprises semantic slots and corresponding values;

the database of frequently-asked queries includes semantic slots and corresponding values for frequently-asked queries; and

the user'"'"'s query is determined to be in the database if the semantic slots and corresponding values in the interpretation result substantially match an entry in the database with a minimum confidence threshold.

13. The computer program of claim 11, further comprising:

for users'"'"' queries not in the database, ranking each type of user query based on the amount of live agent time consumed in handling the query; and

using the rankings, prioritizing queries for adding to the database.

14. The computer program of claim 11, wherein if the user'"'"'s query does not match a query in the database or the user indicates that the user still desires to speak with a live agent, the user is given a high priority for a live agent.

15. The computer program of claim 11, wherein the automated response to the user in accordance with the predetermined response protocol is provided via a voice interface.

16. The computer program of claim 11, wherein the automated response to the user in accordance with the predetermined response protocol is provided via SMS.

17. The computer program of claim 11, wherein the automated response to the user in accordance with the predetermined response protocol is provided via email.

18. The computer program of claim 11, wherein the automated response to the user in accordance with the predetermined response protocol is provided via web chat.

19. The computer program of claim 11, wherein the response protocol identifies any information needed to respond to the query and, if the IVR system does not have all such information, the IVR system'"'"'s initial response to the user is to request more information from the user and wherein, after receiving requested information from the user, the IVR system provides the user with a response to the user'"'"'s query in accordance with the response protocol for the query.