Method and apparatus for analyzing leakage from chat to voice

• US 9,621,725 B2
• Filed: 04/21/2016
• Issued: 04/11/2017
• Est. Priority Date: 01/08/2013
• Status: Active Grant

• Current Assignee: [24]7.ai, Inc.
• Original Assignee: 24/7 Customer Incorporated
• Inventors: Sri, R. Mathangi
• Primary Examiner(s): NGUYEN, QUYNH H
• Application Number:
  US15/135,393
• Publication Number:
  US 20160241714A1
• Time in Patent Office: 355 Days
• Field of Search: 379 9317, 709229
• CPC Class Codes: G06F 40/166 Editing, e.g. inserting or deleting
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06Q 30/016 After-sales
• CPC Class Codes: H04M 2207/203 composed of PSTN and data network, e.g. the Internet
• 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 3/5191 interacting with the Internet
• CPC Class Codes: H04M 7/0045 where the text-based messaging service is an instant messaging service

Abstract

The customer experience is enhanced by detecting leakage-to-voice from chats and providing recommendations to operations, chat agents, and customers. A chat is classified into leakage-to-voice or leakage-to-text chat and actionable recommendations are then provided to operations, chat agents, and customers based on the leakage information. Once leakage is identified, various other insights are extracted from chats and such insights are fed into the knowledge-base. Such insights also used in agent training and are provided to chat agents as recommendations. This results in a better customer experience.

Claims

1. A computer implemented method for analyzing chat leakage, comprising:

providing a processor configured for obtaining chat-related information from at least one chat session conducted over a computer network via a chat communications channel between a customer and an agent;

said processor configured for identifying customer leakage information from said chat session to another communications channel;

said processor configured for building a model based on said chat-related information and said leakage information;

said processor building said model by;

using a chat text to build an anchor;

identifying one or more filters by extracting channel names referred to in the chat text that is used to build said anchor by using a window of words around the anchor to identify a type of channel; and

once the anchors and filters are identified, using a priority matrix to identify an exact channel, wherein said processor is configured for applying said model to identifying a communications channel to which leakage occurs; and

said processor configured for summarizing and passing contextual information of said chat session from said chat communications channel to said another communications channel to avoid repeating collection of said information and to allow said agents to communicate intuitively with said customers to improve the customer experience.

2. The method of claim 1, further comprising:

classifying chats that are identified as having leakage into leakage-to-voice or leakage-to-text chats.

3. The method of claim 1, wherein the contextual information comprises any of:

identity of chat agents, either a voice or a text chat agent, who interacted with the customer before the customer visited a specific chat agent;

a path taken by the customer to reach a chat agent;

product or services related to customer;

customer identity;

the customer'"'"'s Web-log journey; and

other customer information.

4. The method of claim 1, further comprising:

using sample production data to build said model.

5. The method of claim 1, further comprising:

after said anchor is built, using positive hits generated during anchor building in connection with edit-distance to obtain a temporary categorization of a team/department.

6. The method of claim 5, further comprising:

business analysts using business sense for each team/department and customer needs to perform an initial grouping of a portion of teams/departments.

7. The method of claim 5, further comprising:

applying a service layer to fine tune category mapping of said team/department; and

generating a model that includes a final category for each team/department name.

8. The method of claim 7, wherein said category comprises a categories grouping for each team/department name based on edit distance.

9. The method of claim 5, further comprising:

after said anchor is built, categorizing said chat data into said team/department names.

10. The method of claim 9, further comprising:

applying specific filters for voice and text chat categorization.

11. The method of claim 10, wherein said filters and said team/department names occur in a periphery of a predefined number of words;

and

said voice and text chat filters creating dummy data having a periphery of said predefined number of words, instead of using an entire agent chat text; and

providing said dummy data with a dummy identifier.

12. The method of claim 10, further comprising:

providing a priority index to control each of said filters.

13. The method of claim 12, further comprising:

providing a grid editor to provide said priority to said filters.

14. The method of claim 1, wherein said chat text is any of text provided by said customer and by said agent.

15. The method of claim 9, further comprising:

performing a noun phase extraction process in which a part-of-speech (pos) tagger is used to tag lines in the chat based on pos information.

16. The method of claim 15, wherein a first noun from the line of chat text is extracted based on the hypothesis that most of the time the product name is mentioned as the first noun.

17. The method of claim 14, further comprising:

looking up a concordance words with the anchors;

wherein said concordance comprises any of direction (left window/right window/around window) and the window size; and

wherein based on the direction, positive hits are generated by the anchor building process.

18. The method of claim 14, further comprising:

performing a noun phrase extraction process in which a part-of-speech (pos) tagger is used to tag lines in the chat text based on pos information.

19. The method of claim 15, wherein a first noun from the line of chat text is extracted based on the hypothesis that most of the time the product name is mentioned as the first noun.

20. The method of claim 9, further comprising:

performing a stopword removal process to remove stopwords from the chat text.

21. The method of claim 9, further comprising:

performing a get unique word process to remove multiple occurrences of the same team/department names.

22. The method of claim 9, further comprising:

performing a surface-similarity process in which a surface similarity measure is used to determine edit-distance at a word level or at a character level; and

grouping team/department names in a category of the team/department based on similar words.

23. The method of claim 1, further comprising:

passing insights from one channel to a next channel.

24. The method of claim 1, further comprising:

detecting leakage and extracting substantially all contextual and other relevant information from a chat interaction when there is a customer dropout during a chat and the agent cannot solve the customer problem; and

summarizing said relevant information and passing said summarized information to a voice referral.

25. The method of claim 1, further comprising:

adding information regarding leakage to a Web-log journey to predict leakage at the start of a subsequent customer journey when leakage in a current chat is detected; and

providing recommendations to said agent at the start of said subsequent chat, based on leakage detected in said current chat.

26. The method of claim 1, further comprising:

said processor configured for applying said model to provide off line training analysis.