Method and apparatus for extracting journey of life attributes of a user from user interactions

• US 9,910,909 B2
• Filed: 04/10/2014
• Issued: 03/06/2018
• Est. Priority Date: 01/23/2013
• Status: Active Grant

• Current Assignee: [24]7.ai, Inc.
• Original Assignee: 24/7 Customer Incorporated
• Inventors: Hegde, Darshan, Adusumilli, Kranthi Mitra, Kulkarni, Subhash Ramchandra, Prakash, Ravi
• Primary Examiner(s): Gart, Matthew
• Assistant Examiner(s): Swartz, Stephen S
• Application Number:
  US14/250,342
• Publication Number:
  US 20140222823A1
• Time in Patent Office: 1,426 Days
• Field of Search: 705 729, 705 2641, 705319
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data

Abstract

Embodiments of the invention relate to managing user interactions and, more particularly, to performing analysis on data generated by user interactions. Embodiments of the invention use text mining to extract personal information of users from user interactions automatically. A topic model is used to reduce the number of dimensions required to represent the text, yet all the information of interest is highly pronounced. This enables a lower dimensional representation of the data leading to significantly faster computations.

Claims

1. A computer implemented method for performing analysis on data generated by user interactions, comprising:

providing a processor executing instructions for receiving text information from at least one interaction between a user and an agent;

said processor text mining said interaction information to extract personal information relating to said user automatically;

said processor using a topic model to extract lines from said text information to reduce a number of dimensions required to represent the text, wherein all information of interest is highly pronounced, and wherein a resulting lower dimensional representation of the text allows significantly faster computations;

said processor extracting said lines of text as anchored text lines that are indicative of personal information that is present in the anchored text lines;

said processor identifying said anchored text lines by checking for specific keywords which are present when said user is mentioning personal information during said interaction;

said processor applying a statistical technique to said anchored text lines to discover information present in the anchored text lines;

said processor using a k nearest neighbor algorithm to discover said information present in the anchored text lines;

said processor representing said text in a topic space with a score along each axis to indicate an extent to which said text contains personal information about said user;

upon receiving a new data point, which optionally comprises an anchored text line from said text, said processor picking the k closest points to said new data point, determining a predominant class among classes in the k closest points, and assigning said predominant class to said new data point;

said processor assigning a user corresponding to said text to at least one class based on said identified personal information; and

said processor classifying said user based upon said extracted lines of text.

2. The method of claim 1, further comprising:

said processor extracting said interaction information from any of a network and said agent.

3. The method of claim 1, further comprising:

said processor initially transcribing said interaction into text.

4. The method of claim 1, wherein said personal information topic space comprises any of marital status, age, date of birth, travel plans, anniversary date, preferred brands, family related information, financial information, level of education, vehicles owned, location, health related information, level of familiarity with a specific area, and price consciousness.

5. The method of claim 1, further comprising:

said processor gathering data required to identify said personal information from any of said user'"'"'s past interactions, a user profile, and from current interactions.

6. The method of claim 5, further comprising:

said processor continuously generating said user profile by evaluating a plurality of different sets of said data collected across a plurality of channels, multiple data sources, and unique identifiers comprising all of unique data which corresponds to unique identification parameters of the user, aggregate data, transaction data, and said interaction data, said profile including information that uniquely identifies a user as well as the user'"'"'s previous interaction experience and personal information which is used to classify the user.

7. The method of Claim 1, further comprising:

said processor assigning texts a value of ‘

No Segment’

class if personal information about the user cannot be extracted from said text.

8. The method of claim 1, further comprising:

said processor associating an identification with said user.

9. The method of claim 8, wherein said identification comprises any of a user ID associated with the user, a phone number, and an email address.

10. The method of claim 8, further comprising:

once said user has been identified, said processor tailoring interactions with the user based on the identified personal information.

11. An apparatus for performing analysis on data generated by user interactions, comprising:

a processor executing instructions for receiving text information from at least one interaction between a user and an agent;

said processor text mining said interaction information to extract personal information relating to said user automatically;

said processor using a topic model to extract lines from said text information to reduce a number of dimensions required to represent the text, wherein all information of interest is highly pronounced, and wherein a resulting lower dimensional representation of the text allows significantly faster computations;

said processor extracting said lines of text as anchored text lines that are indicative of personal information that is present in the anchored text lines;

said processor identifying said anchored text lines by checking for specific keywords which are present when said user is mentioning personal information during said interaction;

said processor applying a statistical technique said anchored text lines to discover information present in the anchored text lines;

said processor using a k nearest neighbor algorithm to discover said information present in the anchored text lines;

said processor representing said text in a topic space with a score along each axis to indicate an extent to which said text contains personal information about said user;

upon receiving a new data point, which optionally comprises an anchored text line from said text, said processor picking the k closest points to said new data point, determining a predominant class among classes in the k closest points, and assigning said predominant class to said new data point;

said processor assigning a user corresponding to said text to at least one class based on said identified personal information; and

said processor classifying said user based upon said extracted lines of text.