System and method for maximum benefit routing

• US 6,895,083 B1
• Filed: 05/02/2001
• Issued: 05/17/2005
• Est. Priority Date: 05/02/2001
• Status: Expired due to Fees

• Current Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Original Assignee: BBNT Solutions LLC (Rtx Corporation), Verizon Corporate Services Group Incorporated (Verizon Communications Inc.)
• Inventors: Golden, John A., Bers, Joshua, Peterson, Patrick M.
• Primary Examiner(s): Foster, Roland G.
• Assistant Examiner(s): Gauthier, Gerald
• Application Number:
  US09/847,256
• Time in Patent Office: 1,476 Days
• Field of Search: 379/88.01, 379/265.02, 379/265.13, 379/266.02, 379/88.02, 379/88.03, 379/88.04, 379/142.01, 379/142.07, 379/221.01, 379/221.02, 379/221.05, 379/221.06, 379/243, 379/244, 704/240, 704/244, 704/9, 704/249, 704/251, 707/5
• US Class Current: 379/88.01
• CPC Class Codes: H04M 3/5232 Call distribution algorithms
• CPC Class Codes: H04M 3/5233 Operator skill based call distribution
• CPC Class Codes: H04M 3/5237 Interconnection arrangements between ACD systems

Abstract

A system and method for routing telephone calls is based on identifying caller goals, and maximizing the benefit of routing the call to the appropriate call center. The appropriate call center is identified as the call center best equipped to respond to the needs of the caller. The cost or benefit is based on the fastest and least expensive way to answer a query posed by a caller. A probabilistic model of the caller'"'"'s goals or call topics based on a response to a top-level prompt is used, along with a set of functions associating a utility or benefit with routing those call topics to destinations within the center. Maximum benefit routing allows for the separation of the caller'"'"'s goals from those of the call center, thereby optimizing use of available resources.

Claims

1. An automated call routing system that routes a telephone call by responding to a routing objective of a calling party, comprising:

a speech recognizer that determines at least one phrase from a speech utterance made by the calling party and outputs a digital phrase;

a topic identifier that receives the digital phrase and converts the digital phrase to at least one of a word stem and a word class and generates a topic output; and

a maximum benefit router that receives the topic output and determines where to route the telephone call in order to optimize at least one predetermined parameter, said telephone call route based on maximum benefit;

wherein the topic output is a feature vector that contains the number of times the at least one word stems and word classes were found in the determined phrase, and

wherein the maximum benefit router performs at least one of analyzing the feature vector, outputting a posterior possibilities vector, inputting a posterior possibilities vector and determining the expected benefit of routing the call to each of a predetermined destination, and outputting a benefit sorted vector of destinations, benefits and topic scores.

2. The automated call routing system of claim 1, wherein the maximum benefit router determines the best routing objective of the calling party according to call topics.

3. The automated call routing system of claim 1, wherein the maximum benefit router determines the best routing destination based on the routing objective of the calling party distinguished from a second routing objective of a call center.

5. The automated call routing system of claim 1, further comprising a benefit matrix as input to the maximum benefit router, said benefit matrix having at least one routing destination and at least one caller topic.

6. The automated call routing system of claim 1, wherein the topic identifier generates a topic likelihood vector that is input to the maximum benefit router.

8. The automated call routing system of claim 1, wherein the maximum benefit router routes the telephone call to a first call center based upon at least one of optimized time savings, optimized cost savings, optimized response quality and optimized resources.

9. The automated call routing system of claim 1, wherein the maximum benefit router optimizes at least one predetermined parameter using Bayesian decision theory and determining minimum overall risk.

10. The automated call routing system of claim 9, wherein the minimum overall risk is the maximum benefit.

11. The automated call Touting system of claim 1, wherein the speech recognizer is a spoken language understanding device.

4. The automated call routing system of clam 1, wherein the at least one predetermined parameter is selected from an m×

n benefit matrix having m rows and n columns forming an m×

n matrix and where m represents routing destinations and n represents caller topics.

7. The automated call routing system of claim 4, wherein entries in the benefit matrix define the benefit in seconds of agent time saved by routing the call to a first destination based upon a first caller topic.

12. The automated call routing system of clam 1, the topic identifier further comprising a stemming algorithm.

13. The automated call routing system of claim 12, wherein the stemming algorithm is Porter Stemming.

14. An automated call routing system that routes a call by responding to a routing objective of a calling party, comprising:

a recognizer that determines at least one phrase made by the calling party and outputs a second phrase;

a topic identifier that receives the second phrase and converts the second phrase to at least one of a word stem and a word class and generates a topic output; and

a maximum benefit router that receives the topic output and determines where to route the call in order to optimize at least one predetermined parameter, said telephone call routed based on maximum benefit;

wherein the topic output is a feature vector that contains the number of times the at least one word stems and word classes were found in the determined phrase, and

wherein the maximum benefit router performs at least one of analyzing the feature vector, outputting a posterior possibilities vector, inputting a posterior possibilities vector and determining the expected benefit of routing the call to each of a predetermined destination, and outputting a benefit sorted vector of destinations, benefits and topic scores.

15. The automated call routing system of claim 14, wherein the call can be one of a telephone call aud electronic mail.

16. A method for automatically routing a telephone call using maximum benefit routing, comprising the steps of:

receiving a telephone call from a caller;

determining phrases from speech utterances by a caller;

inputting said determined phrases to a speech recognizer device;

converting said recognized determined phrases into at least one of word stems and word classes;

performing keyword lookup on the one of word stems and word classes;

generating a feature vector that contains the number of times the at least one word stems and word classes were found in the determined phrase;

performing analysis on the feature vector;

outputting a posterior possibilities vector;

inputting the posterior possibilities vector and determining the expected benefit of routing the call to each of a predetermined destination; and

outputting a benefit sorted vector of destinations, benefits and topic scores.

17. The method of claim 16, wherein the analysis is performed on the feature vector using one of a multinomial model, a generalized linear model and a support vector machine.

18. The method of claim 17, wherein the posterior possibilities vector is a vector of scores for topics, each score representing confidence that the determined phrase is related to a predetermined topic and vector size is the number of topics.

19. The method of claim 16, further comprising the step of:

determining whether to route the call to a top ranking destination or to reject the utterance if the topic score and/or benefit falls below a predetermined threshold.

20. A maximum benefit call routing system for use in a call center that routes a telephone call made by a calling party by responding to a routing objective of the calling party, comprising:

a speech recognizer that determines at least one phrase from a speech utterance made by the calling party and outputs a digital phrase;

a topic identifier that receives the digital phrase and converts the digital phrase to at least one of a word stem and a word class and generates a topic output; and

a maximum benefit router that receives the topic output and determines where to route the telephone call in order to optimize at least one predetermined parameter, said telephone call routed based on maximum benefit;

wherein the maximum benefit router determines the best routing objective of the calling party according to call topics,

wherein the maximum benefit router determines the best routing destination in the call center based on the routing objective of the calling party distinguished from a second routing objective of the call center,

wherein the topic output is a feature vector that contains the number of times the at least one word stems and word classes were found in the determined phrase, and

wherein the maximum benefit router performs at least one of analyzing the feature vector, outputting a posterior possibilities vector, inputting a posterior possibilities vector and determining the expected benefit of routing the call to each of a predetermined destination, and outputting a benefit sorted vector of destinations, benefits and topic scores.

21. An automated call routing system for use in a call center that routes a call by responding to a routing objective of a calling party, comprising:

a recognizer that determines at least one phrase made by the calling party and outputs a second phrase;

a topic identifier that receives the second phrase and converts the second phrase to at least one of a word stem and a word class and generates a topic output; and

a maximum benefit router that receives the topic output and determines where to route the call in order to optimize at least one predetermined parameter, said telephone call routed based on maximum benefit,

wherein the maximum benefit router determines the best routing objective of the calling party according to call topics,

wherein the maximum benefit router determines the best routing destination in the call center based on the touting objective of the calling party distinguished from a second routing objective of the call center,

wherein the topic output is a feature vector that contains the number of times the at least one word stems and work classes were found in the determined phrase, and

wherein the maximum benefit router performs at least one of analyzing the feature vector, outputting a posterior possibilities vector, inputting a posterior possibilities vector and determining the expected benefit of routing the call to each of a predetermined destination, and outputting a benefit sorted vector of destinations, benefits and topic scores.

22. A method for automatically routing a telephone call in a call center using maximum benefit routing, comprising the steps of:

receiving a telephone call from a caller,

determining phrases from speech utterances by a caller,

inputting said determined phrases to a speech recognizer device;

converting said recognized determined phrases into at least one of word stems and word classes;

performing keyword lookup on the one of word stems and word classes;

generating a feature vector that contains the number of times the at least one word stems and word classes were found in the determined phase;

performing analysis on the feature vector,

outputting a posterior possibilities vector,

inputting the posterior possibilities vector and determining the expected benefit of routing the call to each of a predetermined destination;

outputting a benefit sorted vector of destinations, benefits and topic scores;

inputting the benefit sorted vector and comparing a top ranking destination to a predetermined threshold;

routing the received telephone call to a call center operator if the top ranking destination exceeds the predetermined threshold.