Determining product categories by mining interaction data in chat transcripts

• US 9,460,455 B2
• Filed: 12/31/2013
• Issued: 10/04/2016
• Est. Priority Date: 01/04/2013
• Status: Active Grant

• Current Assignee: [24]7.ai, Inc.
• Original Assignee: 24/7 Customer Incorporated
• Inventors: Hardeniya, Nitin Kumar, Sri, R. Mathangi, Vijayaraghavan, Ravi
• Primary Examiner(s): Cao, Phuong Thao
• Application Number:
  US14/145,570
• Publication Number:
  US 20140195562A1
• Time in Patent Office: 1,008 Days
• Field of Search: 707/776
• US Class Current: 1/1
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06Q 30/01 Customer relationship services
• CPC Class Codes: G06Q 30/0256 User search

Abstract

The propensity and intent of a user to make a purchase is predicted based on product search queries and chat streams. The contents of the data sources, including search queries and chat streams, are analyzed for product names and product attributes. The results of the analyses are used to predict user needs. Product names and attributes are extracted from the data sources. The extracted information is mapped onto abstract product categories. Based on the abstract product categories, offers for products and services are made to the user.

Claims

1. A computer-implemented method for performing analysis of a user'"'"'s interactions, the method comprising:

providing a processor configured for text mining customer interaction data received via one or more user communication channels to establish abstract product categories;

said processor configured for obtaining a query made by the user during a live chat communication with an agent;

said processor configured for extracting keywords from the query that describe a product or a product attribute;

said processor configured for mapping said keywords to one or more of the abstract product categories based on semantic relatedness between the keywords and the abstract product categories;

said processor configured for inferring the user'"'"'s intent based on the one or more abstract product categories;

said processor configured for recommending at least one offer for the user based on the one or more abstract product categories and the user'"'"'s intent; and

said processor configured for updating the abstract product categories in real time as new customer interaction data is received via the one or more user communication channels.

2. The method of claim 1, wherein the one or more user communication channels include any of search queries, web interaction data, and chat streams.

3. The method of claim 1, further comprising:

creating a chat transcript based on the live chat communication,

wherein the chat transcript is a textual representation of a chat session between the user and the agent, and

wherein the abstract product categories are based at least in part on a text mining of the chat transcript.

4. The method of claim 1, further comprising:

a client device communicating with an agent module via a network, wherein the client device is associated with the user and the agent module is associated with the agent,

wherein, during said communication with said agent module, the user, via said client device, discloses information that includes any of certain product related features, product names, and other product relevant information, and

wherein said information is contained in a transcript of a chat session between said user and the agent.

5. The method of claim 1, further comprising providing a Web server that comprises:

a collection module configured for;

collecting a historical data set recorded and received by a search engine, and

collecting keywords, specified categories, and a selection of various aspects tracked by the search engine;

a category determination module configured for managing determination of appropriate categories and query aspects for each of a plurality of categories and for identifying one or more appropriate categories to associate with search queries received by the search engine;

a data processing module configured for processing said historical data set collected by said collection module; and

a database configured for storing said appropriate categories and a top number of key words and tags associated with a search query.

6. The method of claim 3, further comprising:

initially extracting a primary line from said chat transcript, said primary line comprising data regarding an interaction between a client device associated with the user and an agent module associated with the agent;

passing said extracted primary line through a dependency parser which, in turn, outputs a parsed tree; and

based on said parsed tree, implementing a heuristic-based approach to extract a product name or other such relevant information from said chat transcript.

7. The method of claim 3, further comprising:

using a sentence splitter for pre-processing of the chat transcript.

8. The method of claim 3, further comprising:

post processing using any of stop-word removal, spelling correction, and domain specific word removal to obtain cleaner product related information.

9. The method claim 1, said mapping further comprising:

performing product-to-category mapping in which, once a product name and other relevant information is extracted from a chat session, said product name and said relevant information are mapped to an abstract product category to provide insights that are used to produce an actionable recommendation.

10. The method of claim 9, further comprising:

mapping specific domains that comprise entities through an ontology that provides different ontological relations that are used to map a product name, part, and information to product categories.

11. The method of claim 9, further comprising:

mapping a product name to a product category using a family of semantic distance measures, each semantic distance measure being derived from a number of hits returned by a search engine for a given set of keywords,

wherein keywords having a same or similar meaning in a natural language sense are close in units of distance, and

wherein keywords having dissimilar meanings are farther apart; and

using said semantic distance measures to calculate semantic relatedness, wherein said semantic relatedness contributes to a mapping between a product name and a product category.

12. The method of claim 11, further comprising:

using a vocabulary as a metric to calculate semantic relatedness from where it is possible to map a product name to a product category.

13. The method of claim 3, further comprising:

text mining attributes, which include any of promotions, offers, add-on offers, and related parts for a same chat conversation, and any other interaction data between a client device associated with the user and an agent module associated with the agent to create at least one correlation between said attributes,

wherein said at least one correlation is subsequently used by said processor for recommending actions to operations and for agent training.

14. The method of claim 3, further comprising:

a chat client providing access to a knowledge base for agents, wherein said knowledge base recommends offers for a specific product category and related parts for said specific product category.

15. The method of claim 14, wherein said knowledge base provides competitor analysis, offers, and deals for specific product categories.

16. The method of claim 14, further comprising:

using social media data, considering pros and cons for a specific product category, along with sentiment and subjectivity analysis for any of a specific product category, brand, and name.

17. The method of claim 1, further comprising:

making said recommendation of at least one offer at an aggregate level and, when said at least one offer includes a plurality of offers, recommending said plurality of offers offline over time.

18. The method of claim 1, further comprising:

personalizing said recommendation of at least one offer to a specific user visit, wherein each element of real time mining through Web, chat, and social channels generates more personalized recommendations.

19. The method of claim 1, wherein said text mining utilizes a Normalized Google Distance (NGD) function to calculate semantic relatedness, and

wherein said updating of the abstract product categories in real time allows the abstract product categories to change over time.

20. The method of claim 1, further comprising:

said processor configured for using the abstract product categories and the user'"'"'s intent for any of offline analysis, context passing to other channels of interaction, analyzing agent performance, and enhancing the user'"'"'s experience.

21. An apparatus for performing analysis of a user'"'"'s interactions, the apparatus comprising:

a processor; and

a memory that includes instructions, which, when executed by the processor, cause the processor to;

text mine customer interaction data received via one or more communication channels to establish abstract product categories;

obtain a query made by a user during a live chat communication with an agent;

extract keywords from the query that describe a product or a product attribute;

map said keywords to one or more of the abstract product categories based on semantic relatedness between the keywords and the abstract product categories;

infer the user'"'"'s intent based on the one or more abstract product categories;

recommend at least one offer for the user based on the one or more abstract product categories and the user'"'"'s intent; and

update the abstract product categories in real time as new customer interaction data is received via the one or more communication channels.