Real-time call center call monitoring and analysis

• US 9,160,852 B2
• Filed: 11/12/2013
• Issued: 10/13/2015
• Est. Priority Date: 11/21/2012
• Status: Active Grant

• Current Assignee: Castel Communications LLC
• Original Assignee: Castel Communications LLC
• Inventors: Ripa, John, Cheaib, Rachid
• Primary Examiner(s): Hoque, Nafiz E
• Application Number:
  US14/077,990
• Publication Number:
  US 20140140497A1
• Time in Patent Office: 700 Days
• Field of Search: 379/265.02, 379/265.06, 379/265.07, 379/265.12
• US Class Current: 1/1
• CPC Class Codes: G10L 25/48 specially adapted for particular use
• CPC Class Codes: G10L 25/63 for estimating an emotional state
• CPC Class Codes: H04M 2201/12 Counting circuits
• CPC Class Codes: H04M 2201/38 Displays
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/401 Performance feedback
• CPC Class Codes: H04M 3/51 Centralised call answering arrangements requiring operator intervention , e.g. call or contact centers for telemarketing
• CPC Class Codes: H04M 3/5133 Operator terminal details
• CPC Class Codes: H04M 3/5175 Call or contact centers supervision arrangements
• CPC Class Codes: H04M 3/523 with call distribution or queueing

Abstract

Systems and methods are provided for analyzing conversations between customers and call center agents in real-time. An agent may be located at an agent station having a display screen. A continuous audio feed of the conversation between a customer and an agent may be received. For every second that the customer is speaking, a customer emotion score may be calculated in real-time. A frequency at which calculated customer emotion scores equal or exceed an emotion score threshold during a specified time interval may be calculated in real-time during the conversation. The calculated frequency for the customer may be compared, in real-time, to a plurality of specified frequency thresholds. A visual representation corresponding to a highest of the plurality of specified frequency thresholds that is equaled or exceeded by the calculated frequency for the customer may be displayed in real-time on the display screen of the agent station.

Claims

1. A method for analyzing a conversation between a customer and a call center agent in real-time, wherein the agent is located at an agent station having a display screen, the method comprising:

receiving a continuous audio feed of the conversation between the customer and the call center agent;

calculating, in real-time during the conversation, a customer emotion score for every second that the customer is speaking;

calculating, in real-time during the conversation, a frequency at which calculated customer emotion scores equal or exceed an emotion score threshold value during a specified time interval that is less than the entire conversation, wherein calculating the frequency comprises determining a number of calculated customer emotions scores that equal or exceed the emotion score threshold during the specified time interval;

comparing, in real-time during the conversation, the calculated frequency for the customer to a plurality of specified frequency thresholds;

displaying, in real-time on the display screen of the agent station during the conversation, a visual representation corresponding to a highest of the plurality of specified frequency thresholds that is equaled or exceeded by the calculated frequency for the customer;

recording the conversation between the call center agent and the customer;

storing the calculated emotion scores for the customer;

tagging the recorded conversation with event tags, wherein;

each event tag represents the calculated frequency, and

in response to a selection of an event tag, corresponding calculated customer emotion scores are displayed; and

identifying specified emotions and specified word/phrase events that were generated during the recorded conversations.

2. The method of claim 1, further comprising:

generating, in real-time during the conversation, an emotion event whenever the calculated frequency, at which calculated customer emotion scores equal or exceed the emotion score threshold value, equals or exceeds one of the plurality of specified frequency thresholds; and

storing, in the memory, all generated emotion events for the customer.

3. The method of claim 1, further comprising:

calculating, in real-time during the conversation, an agent emotion score for every second that the agent is speaking;

calculating, in real-time during the conversation, a frequency at which calculated agent emotion scores equal or exceed the emotion score threshold during the specified time interval;

comparing, in real-time during the conversation, the calculated frequency for the agent to the plurality of specified frequency thresholds;

displaying, in real-time on the display screen of the agent station during the conversation, a visual representation corresponding to a highest of the plurality of specified frequency thresholds that is equaled or exceeded by the calculated frequency for the agent.

4. The method of claim 3, further comprising:

storing, in a memory, the calculated agent emotion scores;

generating, in real-time during the conversation, an emotion event whenever the calculated frequency, at which calculated agent emotion scores equal or exceed the emotion score threshold value, equals or exceeds one of the plurality of specified frequency thresholds; and

storing, in the memory, all generated emotion events for the agent.

5. The method of claim 3, further comprising displaying, in real-time on a display screen of a supervisor of the agent during the conversation:

the visual representation corresponding to the highest of the plurality of specified frequency thresholds that is equaled or exceeded by the calculated frequency for the customer, and

the visual representation corresponding to the highest of the plurality of specified frequency thresholds that is equaled or exceeded by the calculated frequency for the agent.

6. The method of claim 3, further comprising compiling a report of every customer emotion score, agent emotion score, frequency at which the customer emotion scores equal or exceed the emotion score threshold, and frequency at which the agent emotion scores equal or exceed the emotion score threshold that are calculated during the conversation.

7. The method of claim 1, wherein the plurality of specified frequency thresholds includes at least three frequency thresholds.

8. A method for monitoring conversations at a call center between customers and call center agents in real-time, the method comprising:

calculating emotion scores for agents and respective customers in real-time during

conversations between the agents and the respective customers;

calculating, for each conversation, a first frequency based on the calculated customer emotion scores, wherein calculating the first frequency comprises determining a number of the calculated customer emotions scores that equal or exceed an emotion score threshold during a specified time interval that is less than the entire respective conversation;

calculating, for each conversation, a second frequency based on the calculated agent emotion scores, wherein calculating the second frequency comprises determining a number of the calculated agent emotions scores that equal or exceed the emotion score threshold during the specified time interval;

detecting, in real-time, specified words or phrases spoken by either the agents or the respective customers during the conversations;

displaying, in real-time at an agent station of each agent currently having a conversation with a customer, a visual representation based on the calculated second frequency and a visual representation based on the calculated first frequency;

recording conversations between agents and customers;

storing the calculated emotion scores for agents and customers;

tagging each recorded conversation with event tags, wherein;

each event tag represents one of the first frequency, the second frequency, or a generated word/phrase event, and

in response to a selection of an event tag, corresponding calculated emotion scores and generated word/phrase events are displayed; and

identifying specified emotions and specified additional word/phrase events that were generated during the recorded conversations.

9. The method of claim 8, further comprising:

comparing, in real-time, words or phrases spoken by the agents or the respective customers to at least one keyword list during the conversations;

generating a word/phrase event when a specified word or phrase is detected;

storing, in a memory, the generated word/phrase event; and

displaying, in real-time at an agent station of an agent involved in a conversation during which the specified word or phrase is detected, a message related to the detected word or phrase.

10. The method of claim 8, further comprising generating a report based on the detecting of specified words or phrases, wherein the report comprises a representation of generated word/phrase events and timestamps of when each word/phrase events was generated.

11. The method of claim 8, further comprising detecting, in real-time, targeted data during the conversations between the agents and the respective customers.

12. The method of claim 11, further comprising:

comparing, in real-time, data that is related to the conversations to at least one data dictionary during the conversations in order to detect targeted data;

generating a data event when targeted data is detected;

storing, in a memory, the generated data event; and

displaying, in real-time at an agent station of an agent involved in a conversation during which the specified word or phrase is detected, a message related to the detected targeted data.

13. The method of claim 8, further comprising sending an alert to a supervisor station when calculated customer emotion scores equal or exceed a specified emotion score threshold at a frequency that is equal to or higher than a specified frequency threshold.

14. The method of claim 8, further comprising displaying, at a supervisor station of the call center:

a visual representation based on calculated emotion scores for each of a plurality of agents, and a visual representation based on calculated emotion scores for each of a plurality of respective customers;

a numerical representation of how many word/phrase events have been generated for each of the plurality of agents, and a numerical representation of how many word/phrase events have been generated for each of the plurality of respective customers; and

information corresponding to each respective agent associated with the displayed visual representations and numerical representations.

15. The method of claim 8, wherein a visual representation based on calculated emotion scores of a particular agent is generated based on a frequency at which the calculated emotion scores equal or exceed a specified emotion score threshold, wherein the specified emotion score threshold is set based on personal characteristics of the particular agent.

16. The method of claim 8, further comprising defining a subset of call center agent stations whose respective visual representations based on calculated emotion scores of agents and customers are displayed at a supervisor station, wherein the supervisor station is communicatively coupled to the defined subset of agent stations.

17. The method of claim 8, further comprising:

storing the calculated emotion scores for agents and customers;

identifying past conversations during which all calculated emotion scores have a value below a specified emotion score threshold;

calculating a population mean emotion score by calculating a mean of all calculated emotion scores in the identified past conversations;

calculating, for each agent, an agent mean emotion score by calculating a mean of all calculated emotion scores in all conversations involving a respective agent;

comparing each agent mean emotion score to the population mean emotion score, wherein the comparing comprises calculating a respective variance between each agent mean emotion score and the population mean emotion score; and

generating a report based on the comparing.

18. The method of claim 17, further comprising:

identifying a stored calculated emotion score that is above the specified emotion score threshold as an erroneous calculation; and

replacing the identified stored calculated emotion score with the calculated population mean emotion score.

19. The method of claim 8, further comprising calculating aggregate emotion score statistics and aggregate word/phrase event statistics for conversations involving a particular agent that occurred during a specified time period.

20. A system for analyzing conversations between customers and agents at a call center in real-time, the system comprising:

a multi-threaded audio analyzer for analyzing audio signals on a plurality of customer call channels in real-time, wherein the audio analyzer is configured to;

calculate customer emotion scores in real-time during conversations between customers and agents;

calculate agent emotion scores in real-time during the conversations between customers and agents;

calculate, for each of the conversations between customers and agents, a first frequency based on the calculated customer emotion scores, wherein calculating the first frequency comprises determining a number of the calculated customer emotions scores that equal or exceed an emotion score threshold during a specified time interval that is less than an entire respective conversation;

calculate, for each of the conversations between customers and agents, a second frequency based on the calculated agent emotion scores, wherein calculating the second frequency comprises determining a number of the calculated agent emotions scores that equal or exceed the emotion score threshold during the specified time interval; and

detect, during conversations between customers and agents, whether a plurality of designated words have been spoken;

record conversations between agents and customers;

store the calculated emotion scores for agents and customers;

tag each recorded conversation with event tags, wherein;

each event tag represents one of the first frequency, the second frequency, or a generated word/phrase event, and

in response to a selection of an event tag, corresponding calculated emotion scores and generated word/phrase events are displayed; and

identifying specified emotions and specified additional word/phrase events that were generated during the recorded conversations;

a plurality of agent stations each configured to;

receive audio signals via a respective one of the plurality of customer call channels; and

display, in real-time during a conversation between a respective agent and customer, gradations of color representative of the calculated first frequency and the calculated second frequency;

a supervisor station configured to;

receive audio signals via the plurality of customer call channels;

display, in real-time during conversations between a plurality of agents and customers, gradations of color representative of the calculated first frequency and the calculated second frequency for each of the conversations between customers and agents; and

send instructions to the plurality of agent stations; and

an alert processor for sending alerts to the plurality of agent stations and the supervisor station based on the calculated customer and agent emotion scores and detection of at least some of the plurality of designated words.