Computer systems and methods for efficient query resolution by customer representatives

• US 10,445,745 B1
• Filed: 03/26/2019
• Issued: 10/15/2019
• Est. Priority Date: 03/26/2019
• Status: Active Grant

• Current Assignee: FMR LLC
• Original Assignee: FMR LLC
• Inventors: Chopra, Ankush, Desai, Abhishek, Chandramouli, Aravind
• Primary Examiner(s): Jackson, Jakieda R
• Application Number:
  US16/365,309
• Time in Patent Office: 203 Days
• Field of Search: 704270, 7042701
• CPC Class Codes: G06F 16/31 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06Q 30/016 After-sales

Abstract

A computerized method of representing customer interactions with an organization includes: receiving, by a computing device, customer web interaction data segments and customer conversation data segments; pre-processing the customer conversation data segments to remove specified types of information; scoring each of the pre-processed customer conversation data segments; pre-processing the customer web data interaction segments; extracting from the pre-processed customer web interaction data segments tokens; combining the pre-processed customer conversation data segments and the pre-processed customer web interaction data segments into a customer data set; parsing the customer data set into one or more windows; assigning, for each window, pre-trained weights to each of the tokens in each window; assigning a transaction theme to each window based on the tokens in each window; and generating, based on the transaction themes, a ranked list of topic keywords reflecting the customer web interaction data segments and customer conversation data segments.

Claims

1. A computerized method of representing customer interactions with an organization, the computerized method comprising:

receiving, by a computing device, customer web interaction data segments and customer conversation data segments including at least one of chat data or phone call transcripts;

pre-processing, by the computing device, the customer conversation data segments to remove specified types of information, thereby generating pre-processed customer conversation data segments, each data segment having a timestamp and a customer identifier;

scoring, by the computing device, each of the pre-processed customer conversation data segments using a pre-trained transaction module;

pre-processing, by the computing device, the customer web data interaction segments, thereby generating pre-processed customer web data interaction segments, each data segment having a timestamp and a customer identifier;

extracting, by the computing device, from the pre-processed customer web interaction data segments, one or more tokens;

combining, by the computing device, the pre-processed customer conversation data segments and the pre-processed customer web interaction data segments into a customer data set;

parsing, by the computing device, the customer data set into one or more windows using the timestamps and the customer identifiers, each window including one or more tokens;

assigning, by the computing device, for each window, pre-trained weights to each of the one or more tokens in each window;

assigning, by the computing device, a transaction theme to each window based on the on one or more tokens in each window using an association model; and

generating, by the computing device, based on the transaction themes, a ranked list of topic keywords reflecting the customer web interaction data segments and customer conversation data segments.

2. The method of claim 1 further including displaying, by the computing device, the ranked list of topic keywords in a user interface for a customer service representative.

3. The method of claim 1 further including storing, by the computing device, the ranked list of topic keywords in a database in electronic communication with the computing device.

4. The method of claim 1 wherein pre-processing the customer conversation data segments removes system metadata.

5. The method of claim 1 wherein extracting the one or more keywords from the pre-processed customer conversation data segments includes:

(i) combining, by the computing device, customer interaction data from a single data session into one text file;

(ii) applying, by the computing device, a sentence tokenization algorithm to parse the customer interaction data session into discrete sentence tokens; and

(iii) identifying, by the computing device, key phrases using a stop phrase algorithm to identify, within the discrete sentence tokens, one or more phrases having specified characteristics.

6. The method of claim 1 wherein scoring each of the pre-processed customer interaction data segments includes:

(i) tokenizing, by the computing device, each of the one or more pre-processed customer interaction data segments;

(ii) converting, by the computing device, for each of the one or more pre-processed customer interaction data segments, the tokenized data into a feature-document matrix;

(iii) calculating, by the computing device, a score for each pre-processed customer interaction data segment;

(iv) determining, by the computing device, a highest score; and

(v) assigning, by the computing device, a pre-processed customer interaction data segment having the highest score to the customer conversation data segment, provided that the highest score is above a certain threshold.

7. The method of claim 1 wherein combining the pre-processed customer interaction data segments into a single customer data set includes rolling up the one or more pre-processed customer data segments using customer identifiers and timestamps, thereby creating a chronologically sorted representation of customer interactions for each customer.

8. The method of claim 1 wherein pre-processing the customer web data interaction segments includes at least one of cleaning html encodings from page name tags, adding URL identifiers, or removing channel information.

9. The method of claim 1 wherein the one or more tokens extracted from the pre-processed customer web interaction data segments are at least one of unigram tokens or bigram tokens.

10. The method of claim 1 wherein a new window is created if a customer medium of interaction changes in time or a period of more than a pre-determined time threshold passes with no customer activity occurring.

11. The method of claim 1 wherein assigning pre-trained weights to each of the one or more tokens in each window includes calculating an inverse document frequency (IDF) value for all tokens.

12. The method of claim 1 wherein assigning a transaction theme to each window is based on selecting a top three tokens for each web window.

13. The method of claim 1 wherein generating the ranked list of topic keywords includes applying a formula using context keywords, channel weights, and interaction times as decay weights to score the topic keywords for each customer.

14. A computerized system including:

a computing device including a context engine for representing customer interactions with an organization and a query engine for generating a question-answer set for use by a customer service representative of an organization and mining answers to questions, the computing device configured to;

receive customer web interaction data segments and customer conversation data segments including at least one of chat data or phone call transcripts;

pre-process the customer conversation data segments to remove specified types of information, thereby generating pre-processed customer conversation data segments, each data segment having a timestamp and a customer identifier;

score each of the pre-processed customer conversation data segments using a pre-trained transaction module;

pre-process the customer web data interaction segments, thereby generating pre-processed customer web data interaction segments, each data segment having a timestamp and a customer identifier;

extract, from the pre-processed customer web interaction data segments, one or more tokens;

combine the pre-processed customer conversation data segments and the pre-processed customer web interaction data segments into a customer data set;

parse the customer data set into one or more windows using the timestamps and the customer identifiers, each window including one or more tokens;

assign, for each window, pre-trained weights to each of the one or more tokens in each window;

assign a transaction theme to each window based on the on one or more tokens in each window using an association model; and

generate, based on the transaction themes, a ranked list of topic keywords reflecting the customer web interaction data segments and customer conversation data segments;

one or more databases in electronic communication with the computing device, the one or more databases storing customer web interaction data and customer conversation data including at least one of chat data or phone call transcripts; and

a virtual agent computing device in electronic communication with the computing device.

15. The system of claim 14 further including a chat VA archives database in electronic communication with the computing device;

a historical questions database in electronic communication with the computing device;

a search questions database in electronic communication with the computing device;

a customer demography database in electronic communication with the computing device; and

a user feedback store database in electronic communication with the computing device.