System and method of assigning and managing customer service tickets using natural language processing, machine learning and triviality analysis
A system and method are disclosed for assigning a ticket including one or more databases that store data describing electronic communication between one or more customer system communication devices and one or more service center communication devices. Embodiments further include a computer coupled with one or more databases and configured to monitor communication activity to determine whether a communication channel is opened between one or more customer system communication devices and one or more service center communication devices, responsive to a ticket being opened between one or more customer system communication devices and one or more service center communication devices, analyze the ticket for at least one topic and customer data, search the one or more databases for missing customer data, and in response to detecting missing customer data, identify one or more relationship models that predicts a value for the missing customer data based on known customer data.
1 . A system for assigning a ticket, comprising:
one or more customer systems comprising one or more customer communication devices communicatively connected over a network to one or more service centers comprising one or more service center communication devices, the one or more service center communication devices each corresponding to a particular agent;
a ticket assignment system comprising a server and one or more databases that retrieve data over the network from the one or more service centers and the one or more customer systems and store the data, wherein the data describes electronic communication between the one or more customer communication devices and the one or more service center communication devices;
the server coupled with the one or more databases, the server configured to host one or more runtime processes configured to:
monitor communication activity of the one or more customer communication devices and the one or more service center communication devices to determine whether a ticket has been opened between at least one of the one or more customer communication devices and the one or more service center communication devices;
responsive to a ticket being opened between the at least one of the one or more customer communication devices and the one or more service center communication devices, analyze the ticket for at least one topic and customer data, wherein analyzing the ticket further comprises:
analyze, the communication activity to identify the at least one topic of the communication activity by scoring, by natural language processing, a similarity between the at least one topic and the ticket associated with a customer, calculated as a distance metric between the at least one topic and the ticket by assigning weighted points, and by assigning the weighted points to a matched subject of the ticket, and to ticket data in connection with opening and closing the ticket, and to a matched topic of the ticket and the ticket data;
determine, using the natural language processing, triviality of subsequent one or more communications associated with the closed ticket;
prevent the closed ticket from reopening in response to determining the triviality of the subsequent one or more communications associated with the closed ticket;
search the one or more databases for missing customer data;
in response to detecting the missing customer data, identify one or more relationship models that predict a value for the missing customer data based on known customer data;
train an unsupervised deep learning model, using the one or more relationship models defining relationships between the missing customer data and the known customer data;
predict inferred customer data based on the trained unsupervised deep learning model using confidence scores of the missing customer data, the known customer data, and other inferred customer data;
calculate an agent score for at least two agents for the at least one topic, based on weighted ranking factors of agent skill, customer satisfaction, total time to resolve a closed ticket, and whether the closed ticket was reopened prior to final resolution, wherein the agent score is based, at least in part, on the predicted inferred customer data, wherein the agent score comprises matches of agent data associated with the at least two agents and the predicted inferred customer data;
assign the ticket to a matching agent based, at least in part, on the calculated agent score and a wait time of the matching agent being less than a maximum wait time;
transmit a communication over the network from the ticket assignment system to a service center communication device corresponding to the matching agent;
determine that an accuracy outcome of the matches of the agent data does not improve over time; and
adjust the weighted ranking factors to improve the matches of the agent data.
2 . The system of claim 1 , wherein the agent score is a topic-specific agent score based, at least in part, on the at least one topic and the weighted ranking factors.
3 . The system of claim 1 , wherein the one or more service center communication devices is operably coupled with an electronic display, and the server is further configured to:
display the ticket on one or more service center communication devices associated with the matching agent.
4 . The system of claim 3 , wherein the one or more relationship models predicts a value for the missing customer data based on an inferred customer data score.
5 . A method of assigning a ticket, comprising:
networking one or more customer systems comprising one or more customer communication devices with one or more service centers comprising one or more service center communication devices, the one or more service center communication devices each corresponding to a particular agent;
storing, by a ticket assignment system comprising a server in one or more databases, data retrieved over the network from the one or more service centers and the one or more customer systems that describes electronic communication between the one or more customer communication devices and the one or more service center communication devices;
monitoring, by the server, communication activity of the one or more customer communication devices and the one or more service center communication devices to determine whether a ticket has been opened between at least one of the one or more customer communication devices and the one or more service center communication devices;
in response to a ticket being opened between the at least one of the one or more customer communication devices and the one or more service center communication devices, analyzing the ticket for at least one topic and customer data, wherein analyzing the ticket further comprises:
analyzing, by the server, the communication activity to identify the at least one topic of the communication activity by scoring, by natural language processing, a similarity between the at least one topic and the ticket associated with a customer, calculated as a distance metric between the at least one topic and the ticket by assigning weighted points, and by assigning the weighted points to a matched subject of the ticket, and to ticket data in connection with opening and closing the ticket, and to a matched topic of the ticket and the ticket data;
determining, using the natural language processing, triviality of subsequent one or more communications associated with the closed ticket;
preventing the closed ticket from reopening in response to determining the triviality of the subsequent one or more communications associated with the closed ticket;
searching, by the server, the one or more databases for missing customer data;
in response to detecting the missing customer data, identifying, by the server, one or more relationship models that predicts a value for the missing customer data based on known customer data;
training, by the server, an unsupervised deep learning model, using the one or more relationship models defining relationships between the missing customer data and the known customer data;
predicting, by the server, inferred customer data based on the trained unsupervised deep learning model using confidence scores of the missing customer data, the known customer data, and other inferred customer data;
calculating, by the server, an agent score for at least two agents for the at least one topic, based on weighted ranking factors of agent skill, customer satisfaction, total time to resolve a closed ticket, and whether the closed ticket was reopened prior to final resolution, wherein the agent score is based, at least in part, on the predicted inferred customer data, wherein the agent score comprises matches of agent data associated with the at least two agents and the predicted inferred customer data;
assigning, by the server, the ticket to a matching agent based, at least in part, on the calculated agent score and a wait time of the matching agent being less than a maximum wait time;
transmitting, by the server, a communication over the network from the ticket assignment system to a service center communication device corresponding to the matching agent;
determining that an accuracy outcome of the matches of the agent data does not improve over time; and
adjusting the weighted ranking factors to improve the matches of the agent data.
6 . The method of claim 5 , wherein the agent score is a topic-specific agent score based, at least in part, on the at least one topic and the weighted ranking factors.
7 . The method of claim 5 , further comprising:
displaying the ticket on one or more service center communication devices associated with the matching agent.
8 . The method of claim 7 , wherein the one or more relationship models predicts a value for the missing customer data based on an inferred customer data score.
9 . A non-transitory computer-readable medium embodied with software for assigning a ticket, the software when executed:
networks one or more customer systems comprising one or more customer communication devices with one or more service centers comprising one or more service center communication devices, the one or more service center communication devices each corresponding to a particular agent;
stores, by a ticket assignment system comprising a server, data retrieved over the network from the one or more service centers and the one or more customer systems that describes electronic communication between the one or more customer communication devices and the one or more service center communication devices in one or more databases;
monitors, by the server, communication activity of the one or more customer communication devices and the one or more service center communication devices to determine whether a ticket has been opened between at least one of the one or more customer communication devices and the one or more service center communication devices;
in response to a ticket being opened between the at least one of the one or more customer communication devices and the one or more service center communication devices, analyzes the ticket for at least one topic and customer data, wherein analyzing the ticket further comprises:
analyzes, the communication activity to identify the at least one topic of the communication activity by scoring, by natural language processing, a similarity between the at least one topic and the ticket associated with a customer, calculated as a distance metric between the at least one topic and the ticket by assigning weighted points, and by assigning the weighted points to a matched subject of the ticket, and to ticket data in connection with opening and closing the ticket, and to a matched topic of the ticket and the ticket data;
determine, using the natural language processing, triviality of subsequent one or more communications associated with the closed ticket;
prevent the closed ticket from reopening in response to determining the triviality of the subsequent one or more communications associated with the closed ticket;
searches the one or more databases for missing customer data;
in response to detecting the missing customer data, identifies one or more relationship models that predicts a value for the missing customer data based on known customer data;
trains an unsupervised deep learning model, using the one or more relationship models defining relationships between the missing customer data and the known customer data;
predicts inferred customer data based on the trained unsupervised deep learning model using confidence scores of the missing customer data, the known customer data, and other inferred customer data;
calculates an agent score for at least two agents for the at least one topic, based on weighted ranking factors of agent skill, customer satisfaction, total time to resolve a closed ticket, and whether the closed ticket was reopened prior to final resolution, wherein the agent score is based, at least in part, on the predicted inferred customer data, wherein the agent score comprises matches of agent data associated with the at least two agents and the predicted inferred customer data;
assigns the ticket to a matching agent based, at least in part, on the calculated agent score and a wait time of the matching agent being less than a maximum wait time;
transmits a communication over the network from the ticket assignment system to a service center communication device corresponding to the matching agent;
determine that an accuracy outcome of the matches of the agent data does not improve over time; and
adjust the weighted ranking factors to improve the matches of the agent data.
10 . The non-transitory computer-readable medium of claim 9 , wherein the agent score is a topic-specific agent score based, at least in part, on the at least one topic and the weighted ranking factors.
11 . The non-transitory computer-readable medium of claim 9 , wherein the software when executed further:
displays the ticket on one or more service center communication devices associated with the matching agent.