Predictive Customer Service Environment

• US 20120076283A1
• Filed: 09/21/2011
• Published: 03/29/2012
• Est. Priority Date: 09/23/2010
• Status: Abandoned Application

• Current Assignee: [24]7.ai, Inc.
• Original Assignee: [24]7.ai, Inc.
• Inventors: Vijayaraghavan, Ravi, AJMERA, Dinesh, Panda, Debashish, Ghanshani, Pankai, Kumar, Sumit, Ramachandran, Mathangi Sri
• Application Number:
  US13/239,195
• Publication Number:
  US 20120076283A1
• Time in Patent Office: Days
• US Class Current: 379/93.17
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/006 based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G06Q 30/016 After-sales
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: H04L 12/1827 Network arrangements for conference optimisation or adaptation
• CPC Class Codes: H04L 51/02 using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
• CPC Class Codes: H04L 51/04 Real-time or near real-time messaging, e.g. instant messaging [IM]
• CPC Class Codes: H04M 3/5183 Call or contact centers with computer-telephony arrangements

Abstract

A mechanism for facilitating customer interactions within a customer service environment provides prompt and accurate answers to customer questions. A smart chat facility for use in a customer service environment to predict a customer problem examines a customer chat transcript to identify customer statements that set forth a customer issue and, responsive to this, can route the customer to an agent, an appropriate FAQ, or can implement a problem specific widget in the customer UI. Customer queries are matched with most correct responses and accumulated knowledge is used to predict a best response to future customer queries. The iterative system thus learns from each customer interaction and can adapt to customer responses over time to improve the accuracy of problem prediction.

Claims

1. A computer implemented method for facilitating customer interactions within a customer service environment, comprising:

a processor configured for predicting a customer problem in a customer service environment;

said processor configured for examining a customer chat transcript to identify customer statements that set forth a customer issue; and

said processor configured, responsive thereto, for any of routing said customer to any of an agent and a text-based facility, and for implementing a problem specific widget in a customer UI.

2. The method of claim 1, further comprising:

said processor configured for matching customer queries with most correct responses.

3. The method of claim 2, further comprising:

said processor configured, based upon said matching customer queries with most correct responses, for accumulating knowledge to predict a best response to future customer queries.

4. The method of claim 3, further comprising:

said processor configured for adapting to customer responses over time to improve problem prediction accuracy.

5. An apparatus for problem prediction in a customer service environment, comprising:

an agent console configured for interacting with at least one problem predictor, wherein said problem predictor comprises related articles with which a customer interacts;

a smart chat routing module configured for routing customer chat to service center agents based upon an issue/agent mapping; and

a first module configured for generating real time triggers, said real time triggers comprising any of agent alerts and supervisor alerts, said first module configured for operating in coordination with said agent console;

wherein said triggers are based upon one or more inputs; and

wherein said one or more inputs comprise any of an off line average hold time (AHT) reduction solution and values for customer pre-chat time, issue time, and post-chat time.

6. The apparatus of claim 5, further comprising:

a customer console configured for establishing a smart chat session that is guided by one or more outputs, wherein said one or more outputs comprise any of FAQs that are generated based upon dynamic text mining, machine selection of the right questions to be posed by an agent, and use of appropriate widgets for standard questions.

7. The apparatus of claim 5, further comprising:

a second module configured for implementing an average hold time (AHT) reduction model that is based upon previous and current customer interactions within said service environment, said interaction comprising any of customer interactions with any of agent-based, text-based, and widget-based assistance.

8. The apparatus of claim 5, wherein chat related information is divided into pre-chat, actual chat, and post chat.

9. A computer implemented method for facilitating customer interactions within a customer service environment, comprising:

a processor configured for examining a customer chat transcript to identify customer statements that set forth a customer issue;

said processor configured for identifying a position in said chat transcript at which an issue is posed by the customer;

said processor configured for text mining said chat transcript for key lines that identify the customer issue;

said processor configured for dividing time components within said chat transcript between agent time and customer time, wherein said time components comprise any of pre-chat, actual chat, and post chat;

said processor configured, as a result of said examining, identifying, text mining, and dividing operations, for producing a dataset, said dataset comprising any of customer ID, main issue, one or more sub-issues, agent/customer time for a main issue, agent/customer time for sub-issues, transfer type, transfer time, hold time, pre-chat time, post-chat time, number of customer lines, number of agent lines, number of lines of separation, and status concerning disconnections by customer chats.

10. A computer implemented method for facilitating customer interactions within a customer service environment, comprising:

a processor configured for extracting a primary question during a chat session by identifying a first question in a chat session transcript and identifying a position of the question in the chat transcript;

wherein if the position of the question in the transcript is greater than or equal to a predetermined value, then the issue is extracted;

else, the transcript is analyzed to determine if a predetermined phrase is present;

wherein if said predetermined phrase if found, then a next sentence is deemed to be the primary question;

else, the identifies as the primary question a first customer sentence in the transcript that has a predetermined number of words.

11. The method of claim 10, further comprising:

said processor configured for, after the primary question has been extracted, extracting an issue by extracting a list of text-based materials with regard to the primary question, said list comprising one or more related answers;

said processor configured for extracting unigrams and bigrams from the primary question;

said processor configured for preparing a list of top unigrams and bigrams based upon their occurrence in customer issues;

said processor configured for mapping said bigrams with said customer issues;

said processor configured for mapping said unigrams and bigrams back to one or more query categories;

wherein, if there is a match, then the issue has been successfully extracted;

else, a match is made to the unigram to obtain the customer issue.

12. An apparatus for facilitating customer interactions within a customer service environment, comprising:

a processor configured for predicting a customer problem in a customer service environment;

said processor configured for examining a customer chat transcript to identify customer statements that set forth a customer issue; and

said processor configured, responsive thereto, for any of routing said customer to any of an agent and a text-based facility, and for implementing a problem specific widget in a customer UI.

13. The apparatus of claim 12, further comprising:

said processor configured for matching customer queries with most correct responses.

14. The apparatus of claim 13, further comprising:

said processor configured, based upon said matching customer queries with most correct responses, for accumulating knowledge to predict a best response to future customer queries.

15. The apparatus of claim 14, further comprising:

said processor configured for adapting to customer responses over time to improve problem prediction accuracy.

16. A computer implemented method for problem prediction in a customer service environment, comprising:

providing an agent console configured for interacting with at least one problem predictor, wherein said problem predictor comprises related articles with which a customer interacts;

providing a smart chat routing module configured for routing customer chat to service center agents based upon an issue/agent mapping; and

providing a first module configured for generating real time triggers, said real time triggers comprising any of agent alerts and supervisor alerts, said first module configured for operating in coordination with said agent console;

wherein said triggers are based upon one or more inputs; and

wherein said one or more inputs comprise any of an off line average hold time (AHT) reduction solution and values for customer pre-chat time, issue time, and post-chat time.

17. The method of claim 16, further comprising:

providing a customer console configured for establishing a smart chat session that is guided by one or more outputs, wherein said one or more outputs comprise any of FAQs that are generated based upon dynamic text mining, machine selection of the right questions to be posed by an agent, and use of appropriate widgets for standard questions.

18. The methods of claim 16, further comprising:

providing a second module configured for implementing an average handle time (AHT) reduction model that is based upon previous and current customer interactions within said service environment, said interaction comprising any of customer interactions with any of agent-based, text-based, and widget-based assistance.

19. The method of claim 16, wherein chat related information is divided into pre-chat, actual chat, and post chat.

20. An apparatus for facilitating customer interactions within a customer service environment, comprising:

a processor configured for examining a customer chat transcript to identify customer statements that set forth a customer issue;

said processor configured for identifying a position in said chat transcript at which an issue is posed by the customer;

said processor configured for text mining said chat transcript for key lines that identify the customer issue;

said processor configured for dividing time components within said chat transcript between agent time and customer time, wherein said time components comprise any of pre-chat, actual chat, and post chat;

said processor configured, as a result of said examining, identifying, text mining, and dividing operations, for producing a dataset, said dataset comprising any of customer ID, main issue, one or more sub-issues, agent/customer time for a main issue, agent/customer time for sub-issues, transfer type, transfer time, hold time, pre-chat time, post-chat time, number of customer lines, number of agent lines, number of lines of separation, and status concerning disconnections by customer chats.

21. An apparatus for facilitating customer interactions within a customer service environment, comprising:

a processor configured for extracting a primary question during a chat session by identifying a first question in a chat session transcript and identifying a position of the question in the chat transcript;

wherein if the position of the question in the transcript is greater than or equal to a predetermined value, then the issue is extracted;

else, the transcript is analyzed to determine if a predetermined phrase is present;

wherein if said predetermined phrase if found, then a next sentence is deemed to be the primary question;

else, the identifies as the primary question a first customer sentence in the transcript that has a predetermined number of words.

22. The apparatus of claim 21, further comprising:

said processor configured for, after the primary question has been extracted, extracting an issue by extracting a list of text-based materials with regard to the primary question, said list comprising one or more related answers;

said processor configured for extracting unigrams and bigrams from the primary question;

said processor configured for preparing a list of top unigrams and bigrams based upon their occurrence in customer issues;

said processor configured for mapping said bigrams with said customer issues;

said processor configured for mapping said unigrams and bigrams back to one or more query categories;

wherein, if there is a match, then the issue has been successfully extracted;

else, a match is made to the unigram to obtain the customer issue.

23. An apparatus for event-driven, customizable action execution to facilitate contextual interactions, comprising:

a processor configured as a predictive service platform for building and provisioning real time interaction management solutions over a network;

said processor configured to capture information representative of a user'"'"'s journey across said network;

said processor configured to model said journey as a finite state machine consisting of distinct states and conditional transitions between them; and

a plurality of different event handlers configured to take specific actions in response to said distinct states and conditional transitions between them.

24. The apparatus of claim 23, wherein a page load results into a state transition, any event can cause a state transition, and all metadata for action invocation is attached to a state.

25. The apparatus of claim 23, said information comprising any of data, actions, and time on all pages.

26. The apparatus of claim 23, wherein an event handler can be configured for any of:

causing state transition;

showing up an interaction popup;

providing a self service wizard, or any other customized interaction interface to a user; and

initiating a chat conversation at any point in time.

27. The apparatus of claim 23, said processor configured as a predictive model for predicting issues and resolutions.

28. The apparatus of claim 23, wherein multiple mediums of interaction can be employed during an interaction, including any of chat, self service, emails, social media, and click to call.

29. The apparatus of claim 23, said processor configured to generate one or more sessions to identify any of a visitor, a logical browsing session of the user, and a logical interaction with the user.

30. The apparatus of claim 23, said processor configured to send periodical and on-demand update of tracked information to a server.

31. The apparatus of claim 23, wherein said actions comprise determining any of:

whom to engage based on context, including collected data elements of a visitor;

when and where to engage based on context, including the journey of the user and time spent on the journey; and

what to show based on context, including a targeted way to engage a customer.

32. An apparatus for event-driven, customizable action execution to facilitate contextual interactions, comprising:

a processor configured as a predictive service platform for building and provisioning real time interaction management solutions over a network;

said processor configured to capture information representative of a user'"'"'s journey across said network; and

said processor configured to generate events during a lifecycle of journey responsive to said captured information.

33. The apparatus of claim 32, said processor configured to model said journey as a finite state machine consisting of distinct states and conditional transitions between them;

and

further comprising a rules engine configured to divide decision making into three phases, comprising;

client side data collection and client side condition evaluation;

server side data collection from third party integrations; and

server side condition evaluation based on client, as well as server side, data.

34. The apparatus of claim 32, said processor configured to allow other system components to subscribe for, and take appropriate actions.

35. The apparatus of claim 33, wherein client side data comprises any of:

time on a page;

geography;

cookies;

DOM data;

client side persistent storage; and

data obtained anywhere during a journey.

36. The apparatus of claim 33, wherein server side data comprises any of:

a page visitor'"'"'s profile; and

past history and any third party data coming from a backend.

37. An apparatus, comprising:

a processor configured to execute a program that guides a user through a resolution path;

said processor configured to capture information representative of said user'"'"'s journey through said resolution path to track an exact state of the user and for storing said information in a persistent local storage medium;

said processor configured to provide a synchronization construct for multiple browser tabs and/or windows to access said storage medium in a consistent manner;

wherein said storage medium is enabled by said processor to allow a stateless connection to read from and write to said multiple browser tabs and/or windows.

38. The apparatus of claim 37, said processor configured to effect read/write synchronization on said storage medium by dividing said storage medium into different areas and using a browser'"'"'s events to store said information into said storage medium from said multiple browser windows and/or tabs.

39. The apparatus of claim 38, said events comprising any of a focus event, blur event, mouse over event, unload event, and onload event.

40. The apparatus of claim 38, said processor configured to create at least two logical storage areas to store all said information, including a shared storage area which is common for a Website and for all tabs and windows and is accessed, read and write, from all a windows;

and a window storage area which every window has its own copy of and which accessed, read only, from all windows.

41. The apparatus of claim 37, further comprising:

said processor configured to implement a locking mechanism to provide a given window a right to read or write into a shared storage area.

42. The apparatus of claim 41, wherein said locking mechanism is implemented using a combination of focus, blur, and mouse over events generated by a browser;

and lock information, including timestamp, that is maintained inside a window storage area of each window.

43. The apparatus of claim 42, wherein said events switch said locking mechanism from one window to another;

wherein lock information stored in said window storage area determines which window is currently locked.