Conversation analysis of asynchronous decentralized media
View Patent ↗The present disclosure provides a system that allows for the real-time and online monitoring of the exchanges between customers and a CRM team over social media. While crawling all messages exchanged over the social media by customers and CRM team, the system aggregates related messages exchanged between a given customer and the CRM team into a conversation. The system includes a linguistic framework for the analysis of conversations (based on the two linguistic theories of dialog acts and conversation analysis) to label the nature of the messages in a conversation or thread.
1. A method for computing online metrics for customer relations management, comprising:
querying historical linguistic data sets of exchanges between customers and customer relations representatives over social media;
extracting from the historical linguistic data sets conversations between said customers and said customer relations representatives;
wherein extracting conversations uses a customer relations representative identifier and extracts all messages sent by said customer relations representative identifier or to said customer relations representative identifier;
wherein said messages are grouped by a customer using a customer associated identifier;
ordering chronologically a list of all said messages by said customer associated identifier;
extracting from the list of all said messages groups of messages, wherein each group of messages is related to an individual subject;
aggregating said groups of messages into pseudo synchronous conversations;
separating each pseudo synchronous conversation of the pseudo synchronous conversations into one or more segments and associating one or more classes to each of the one or more segments;
labeling each said segment in said each pseudo synchronous conversation according to a conversation analysis; and,
labeling each said segment in said each pseudo synchronous conversation according to an engagement analysis.
2. The method of claim 1 , wherein said data sets are asynchronous and decentralized.
3. The method of claim 1 , further comprising:
obtaining as many lists as there are customers.
4. The method of claim 1 , wherein if the lapsed time between consecutive messages is greater than or equal to three days, then identify pseudo conversation as two different conversations.
5. The method of claim 1 , wherein if the lapsed time between consecutive messages is less than three days, then identify pseudo conversation as same conversation.
6. The method of claim 1 , wherein each said labeling according to conversation analysis is selected from the group consisting of: complaint, apology, answer, receipt, compliment, response to positivity, request, greeting, thank, announcement, solved, and other.
7. The method of claim 1 , wherein each said labeling according to said engagement analysis is selected from the group consisting of: open, solved, closed, and change channel.
8. The method of claim 1 , further comprising:
producing metrics from the segments, wherein said metrics are selected from the group consisting of: message thread length, time thread length, first wait time, average wait time, and arrival rate.
9. The method of claim 1 , further comprising:
computing engagement metrics from said labels of each said segment, wherein said engagement metrics are selected from the group consisting of: resolution rate, properly handled threads rate, customer hang-up, conversion rate, and happy customer rate.
10. A method comprising:
querying historical linguistic data sets of exchanges between customers and customer relations representatives over social media;
extracting conversations from the historical linguistic data sets between said customers and said customer relations representatives;
wherein extracting conversations uses a customer relations representative identifier and extracts all messages sent by said customer relations representative identifier or to said customer relations representative identifier;
wherein said messages are grouped by a customer using a customer associated identifier;
ordering chronologically a list of all said messages by said customer associated identifier;
extracting from the list of all said messages groups of messages, wherein each group of messages is related to an individual subject;
aggregating said groups of messages into pseudo synchronous conversations;
separating each pseudo synchronous conversation of the pseudo synchronous conversations into one or more segments and associating one or more classes to each of the one or more segments;
labeling each segment in said each pseudo synchronous conversation according to conversation analysis; and,
computing engagement metrics from said labels of each said segment, wherein said engagement metrics are selected from the group consisting of: resolution rate, properly handled threads rate, customer hang-up, conversion rate, and happy customer rate.
11. The method of claim 10 , wherein said data sets are asynchronous and decentralized.
12. The method of claim 10 , further comprising:
obtaining as many lists as there are customers.
13. The method of claim 10 , wherein if the lapsed time between consecutive messages is greater than or equal to three days, then identify pseudo conversation as two different conversations.
14. The method of claim 10 , wherein if the lapsed time between consecutive messages is less than three days, then identify pseudo conversation as same conversation.
15. The method of claim 10 , wherein each said labeling according to engagement analysis is selected from the group consisting of: open, solved, closed, and change channel.
16. The method of claim 10 , further comprising:
producing metrics from the segments, wherein said metrics are selected from the group consisting of: message thread length, time thread length, first wait time, average wait time, and arrival rate.
17. The method of claim 10 , wherein each said labeling according to conversation analysis is selected from the group consisting of: complaint, apology, answer, receipt, compliment, response to positivity, request, greeting, thank, announcement, solved, and other.
18. A non-transitory machine-readable storage medium having embodied thereon instructions executable by one or more machines to perform operations, comprising:
querying historical linguistic data sets;
extracting from the historical linguistic data sets conversations between customers and customer relations representatives;
wherein extracting the conversations uses a customer relations representative identifier and extracts all messages sent by said customer relations representative identifier or to said customer relations representative identifier;
wherein said messages are grouped by a customer using a customer associated identifier;
ordering chronologically a list of all said messages by said customer associated identifier;
extracting from the list of all said messages groups of messages, wherein each group of messages is related to an individual subject;
aggregating said groups of messages into pseudo synchronous conversations;
separating each pseudo synchronous conversation of the pseudo synchronous conversations into one or more segments and associating one or more classes to each of the one or more segments;
labeling each said segment in said each pseudo synchronous conversation according to a conversation analysis;
labeling each said segment in said each pseudo synchronous conversation according to an engagement analysis;
wherein said historical linguistic data sets are asynchronous and decentralized; and,
wherein if a lapsed time between consecutive messages is greater than or equal to a predeterminable time period, then identify a pseudo conversation as two different conversations.
19. The method of claim 18 , further comprising:
computing engagement metrics from said labels of each said segment, wherein said engagement metrics are selected from the group consisting of: resolution rate, properly handled threads rate, customer hang-up, conversion rate, and happy customer rate.
20. The method of claim 19 , wherein each said labeling according to said engagement analysis is selected from the group consisting of: open, solved, closed, and change channel.