Disclosure of Invention
The technical task of the invention is to provide an AI-based call center management platform which integrates the subjective answering service score of the customer and the objective parameters of AI answering call return as evaluation basis aiming at the problems existing in the background technology, and can effectively solve the problems mentioned in the background technology.
The invention is realized by the following technical scheme:
a call center management platform based on AI comprises a target AI answering port determining module, an answering call record extracting module, an answering call voice matching degree collecting module, an answering call timeliness collecting module, an answering call relevancy collecting module, a call ending service grading collecting module, a modeling processing module, a management database, a management server and a display terminal;
the target AI answering port determining module is used for determining a target AI answering port corresponding to a current calling client when a client call is received;
the answer call record extraction module is used for extracting the whole answer call record of the current calling client answered by the target AI answer port when the current client calls;
the answering and answering voice matching degree acquisition module is used for processing the extracted answering call records and acquiring the answering voice parameter matching degree corresponding to the target AI answering port, wherein the answering and answering voice matching degree acquisition module comprises a voice parameter extraction unit and a answering voice parameter matching degree analysis unit of the two parties;
the answer call-back timeliness acquisition module is used for processing the extracted answer call record and acquiring the call-back timeliness of the target AI answer port for replying the problem proposed by the client;
the answer call return association degree acquisition module is used for processing the extracted answer call records and acquiring the call return association degree of a target AI answer port replying a question proposed by a client;
the call completion service score acquisition module is used for acquiring answering service scores of the target AI answering port by the client after answering the call;
the modeling processing module is used for evaluating the comprehensive answering service quality coefficient of the target AI answering port to the current calling client by combining the answering voice parameter matching degree corresponding to the target AI answering port, the answering timeliness of the target AI answering port for answering the question proposed by the client, the answering relevance of the target AI answering port for answering the question proposed by the client and the answering service score of the client to the target AI answering port;
the management server is used for numbering all AI answering ports in the AI calling center, counting the total times of answering client calling corresponding to all AI answering ports in a set time period, and further evaluating the comprehensive answering service quality coefficient corresponding to all the AI answering ports in the set time period according to the evaluation method of the comprehensive answering service quality coefficient of the target AI answering port to the current calling client, so that the AI answering ports with poor answering service quality in the set time period are identified according to the evaluation result;
the display terminal is used for displaying the number corresponding to the AI answering port with poor answering service quality.
Preferably, the specific method for determining the target AI answering port corresponding to the current calling client is to acquire the current answering state corresponding to each AI answering port in the AI calling center, further screen out the AI answering ports in the missed answering state, determine the AI answering ports as the AI answering ports corresponding to the current calling client, and mark the AI answering ports as the target AI answering ports.
Preferably, the speech parameters include speech rate and pitch.
Preferably, the two-party speech parameter extraction unit is used for extracting the speaking speech of the client and the return speech of the target AI answering port from the extracted answering call records, further extracting the speaking speech parameters of the client from the speaking speech of the client, and extracting the return speech parameters from the return speech of the target AI answering port.
Preferably, the matching degree analysis unit of the answering speech parameters is used for analyzing the matching degree of the answering speech parameters of the target AI answering port, and the specific analysis process comprises the following steps:
s1, comparing the speaking voice parameter of the client with the proper answer voice parameter of the AI answer port corresponding to various speaking voice parameters of the client preset in the management database to obtain the proper answer voice parameter of the AI answer port corresponding to the speaking voice parameter of the client;
s2, the return speech parameters of the target AI listening port and the speech parameters of the clientThe corresponding AI answering port is suitable for matching the return voice parameters, and the matching degree of the return voice parameters corresponding to the target AI answering port is counted
Sigma is expressed as the matching degree of the echo voice parameters corresponding to the target AI listening port, v and u are expressed as the echo speed and the echo tone of the target AI listening port respectively, v and u are expressed as the echo tone of the target AI listening port respectively
0、u
0Respectively representing the suitable answering speech speed and the suitable answering tone of the AI answering port corresponding to the speaking speech parameter of the client.
Preferably, the answer call and timeliness acquisition module acquires a specific acquisition process corresponding to the answer call timeliness of the target AI answer port replying the problem proposed by the client as follows:
h1, counting the number of questions proposed by the client from the extracted answering call records, numbering the questions proposed by the client according to the sequence of the proposing time, and marking the questions as 1,2, a.
H2, acquiring a proposal ending time point corresponding to each question put forward by the client and a call returning time point for replying each question by the target AI answering port from the extracted answering call records;
h3, subtracting the time point of answering the question from the time point of answering the question by the target AI answering port to obtain the answering time of the question by the target AI answering port, and recording as tk;
H4, comparing the call back time of each question replied by the target AI answering port with the predefined call back time, and counting the call back time of the target AI answering port replying the question raised by the client
Eta represents the call back timeliness, t, for the target AI answering port to answer the client's proposed question
0Expressed as a predefined callback duration.
Preferably, the answer call relevancy collection module collects the answer call relevancy corresponding to the answer call relevancy of the target AI answer port replying the problem proposed by the client as follows:
d1, intercepting the question voice information of each question proposed by the client from the extracted answering call record;
d2, extracting question keywords from the intercepted question voice information, wherein the specific extraction method comprises the following steps:
d21, performing question text recognition on the question voice information of each question;
d22, extracting the question key words of the question texts corresponding to the questions;
d3, intercepting the return voice information of each question proposed by the target AI answering port to the client from the extracted answering voice call records;
d4, extracting the answer key words of the intercepted answer voice information to obtain answer key words corresponding to each question extracted by the answer client of the target AI answer port;
d5, matching the answer key words corresponding to the questions proposed by the target AI answering port answering client with the question key words corresponding to the questions, if the answer key words corresponding to the questions proposed by the target AI answering port answering client are successfully matched with the question key words corresponding to the questions, the answer correlation index corresponding to the questions is recorded as epsilon, otherwise, the answer correlation index corresponding to the questions is recorded as epsilon';
d6, according to the answer correlation index corresponding to each question proposed by the target AI answering port answer client, counting the answer correlation degree of the question proposed by the target AI answering port answer client
Xi represents the degree of call-back association, λ, for the target AI listening port to answer the customer's proposed question
kThe answer correlation index, lambda, corresponding to the kth question is provided for the answer client of the target AI answering port
kThe value of (d) may be epsilon or epsilon'.
Preferably, the evaluation calculation formula of the target AI answering port for the comprehensive answering service quality coefficient of the current calling client is
Expressed as the comprehensive answering service quality coefficient of the target AI answering port to the current calling client, q is expressed as the answering service grade value of the client to the target AI answering port
0Expressed as the highest scoring value of the call service, α 1, α 2, α 3, α 4 are expressed as the evaluation weight scaling coefficients of the answering voice parameter matching degree, the answering timeliness, the answering relevance degree, the answering service score to the comprehensive answering service quality coefficient, and α 1+ α 2+ α 3+ α 4 is 1.
Preferably, the size relationship corresponding to α 1, α 2, α 3, α 4 is α 3> α 2> α 4> α 1.
Preferably, the specific identification process for identifying the AI listening port with poor listening service quality within a set time period according to the evaluation result includes the following steps:
f1, comparing the comprehensive answering service quality coefficient corresponding to each answering client call in the set time period of each AI answering port with the standard comprehensive answering service quality coefficient set in the management database, and counting the times of answering client calls of each AI answering port in the set time period which are less than the set standard comprehensive answering service quality coefficient;
f2, dividing the total times of customer calls answered by each AI answering port in a set time period by the times of customer calls answered by the AI answering port in the set time period which is less than the set standard comprehensive answering service quality coefficient to obtain the corresponding answering service bad ratio coefficient of each AI answering port in the set time period;
f3, comparing the poor ratio coefficient of the answering service corresponding to each AI answering port in the set time period with the set value, if the poor ratio coefficient of the answering service corresponding to a certain AI answering port in the set time period is larger than the set value, the AI answering port is marked as the poor quality AI answering port.
The invention has the following beneficial effects:
(1) the invention determines the target AI answering port corresponding to the current calling client when receiving the client call, extracts the whole answering call record of the current calling client answered by the target AI answering port, processes the extracted answering call record, collects the matching degree of the answering voice parameter corresponding to the target AI answering port, the answering timeliness of the answering voice of the target AI answering port for answering the problem posed by the client and the answering relevance of the answering service of the target AI answering port for answering the problem posed by the client, and collects the answering service score of the client for the target AI answering port after the answering call is finished, thereby comprehensively evaluating the comprehensive answering service quality coefficient of the target AI answering port for the current calling client, realizing the comprehensive evaluation of the AI answering service quality according to the combination of the subjective answering service score of the client and the objective answering call parameter of the AI, the evaluation sidedness caused by the evaluation of the existing evaluation mode of the answering service quality of the AI call center only according to the subjective answering service score of the customer is avoided, so that the reliability of the evaluation result is improved.
(2) The invention obtains the comprehensive answering service quality coefficient corresponding to each time of answering the client call in the set time period by adopting the evaluation method of the target AI answering port to the current calling client comprehensive answering service quality coefficient, and further identifies the AI answering port with poor answering service quality in the set time period according to the evaluation result.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, an AI-based call center management platform includes a target AI answering port determining module, an answering call record extracting module, an answering call voice matching degree collecting module, an answering call timeliness collecting module, an answering call relevancy collecting module, a call ending service score collecting module, a modeling processing module, a management database, a management server, and a display terminal.
The target AI answering port determining module is connected with the answering call record extracting module, the answering call record extracting module is respectively connected with the answering call voice matching degree collecting module, the answering call timeliness collecting module and the answering call relevancy collecting module, the answering call voice matching degree collecting module, the answering call timeliness collecting module, the answering call relevancy collecting module and the call ending service grading collecting module are all connected with the modeling processing module, the modeling processing module is connected with the management server, and the management server is connected with the display terminal.
The specific determination method comprises the steps of obtaining current answering states corresponding to all AI answering ports in an AI calling center, further screening the AI answering ports in the un-answering state from the current answering states, determining the AI answering ports as the AI answering ports corresponding to the current calling clients, and recording the AI answering ports as the target AI answering ports.
The answering call record extraction module is used for extracting the whole answering call record of the current calling client answered by the target AI answering port when the current client calls.
Referring to fig. 2, the answering call voice matching degree acquisition module is configured to process the extracted answering call record, and acquire a call return voice parameter matching degree corresponding to the target AI answering port from the answering call record, where the voice parameters include speed and pitch, and the answering call voice matching degree acquisition module includes a two-party voice parameter extraction unit and a call return voice parameter matching degree analysis unit.
The two-party voice parameter extraction unit is used for respectively extracting the speaking voice of the client and the return voice of the target AI answering port from the extracted answering call records, further extracting the speaking voice parameter of the client from the speaking voice of the client and extracting the return voice parameter from the return voice of the target AI answering port.
The return speech parameter matching degree analysis unit is used for carrying out matching degree analysis on the return speech parameters of the target AI answering port, and the specific analysis process executes the following steps:
s1, comparing the speaking voice parameter of the client with the proper answer voice parameter of the AI answer port corresponding to various speaking voice parameters of the client preset in the management database to obtain the proper answer voice parameter of the AI answer port corresponding to the speaking voice parameter of the client;
s2, matching the answering speech parameter of the target AI answering port with the suitable answering speech parameter of the AI answering port corresponding to the speaking speech parameter of the client, and counting the matching degree of the answering speech parameter corresponding to the target AI answering port
Sigma is expressed as the matching degree of the echo voice parameters corresponding to the target AI listening port, v and u are expressed as the echo speed and the echo tone of the target AI listening port respectively, v and u are expressed as the echo tone of the target AI listening port respectively
0、u
0The suitable answering speech speed and the suitable answering tone of the AI answering port corresponding to the speaking speech parameter of the client are respectively expressed, wherein the closer the answering speech parameter of the target AI answering port is to the suitable answering speech parameter of the AI answering port corresponding to the speaking speech parameter of the client, the greater the matching degree of the answering speech parameter is.
In the embodiment, the acquisition of the matching degree of the call-back voice parameters corresponding to the target AI answering port intuitively reflects the matching degree of the call-back voice parameters and the speaking voice parameters of the client in the answering call record of the target AI answering port, and provides the relevant coefficient of the matching of the call-back voice parameters for the evaluation of the comprehensive answering service quality coefficient.
The answering call timeliness acquisition module is used for processing the extracted answering call records and acquiring the call timeliness of the target AI answering port for replying the problem proposed by the client, and the specific acquisition process is as follows:
h1, counting the number of questions proposed by the client from the extracted answering call records, numbering the questions proposed by the client according to the sequence of the proposing time, and marking the questions as 1,2, a.
H2, acquiring a proposal ending time point corresponding to each question put forward by the client and a call returning time point for replying each question by the target AI answering port from the extracted answering call records;
h3, subtracting the time point of answering the question from the time point of answering the question by the target AI answering port to obtain the answering time of the question by the target AI answering port, and recording as tk;
H4, comparing the call back time of each question replied by the target AI answering port with the predefined call back time, and counting the call back time of the target AI answering port replying the question raised by the client
Eta represents the call back timeliness, t, for the target AI answering port to answer the client's proposed question
0And the time is expressed as a predefined call-back time, wherein the closer the call-back time for the target AI listening port to answer each question is to the predefined call-back time, the greater the call-back time.
In the embodiment, the acquisition of the timeliness of the question answering for the reply client corresponding to the target AI answering port intuitively reflects the timeliness of the question answering for the reply client in the answering call record of the target AI answering port, and provides a correlation coefficient of the timeliness of the question answering for the evaluation of the comprehensive answering service quality coefficient.
The answering call relevance degree acquisition module is used for processing the extracted answering call records and acquiring the call relevance degree of a target AI answering port replying a problem proposed by a client, and the specific acquisition process is as follows:
d1, intercepting the question voice information of each question proposed by the client from the extracted answering call record;
d2, extracting question keywords from the intercepted question voice information, wherein the specific extraction method comprises the following steps:
d21, performing question text recognition on the question voice information of each question;
d22, performing stop word and segmentation processing on the problem text corresponding to each problem, and further extracting problem keywords from the obtained segmentation;
d3, intercepting the return voice information of each question proposed by the target AI answering port to the client from the extracted answering voice call records;
d4, extracting the return speech keyword of the intercepted return speech information, wherein the specific extraction method comprises the following steps:
d41, recognizing the return speech text of the return speech information of each question;
d42, carrying out stop word and segmentation processing on the answer text corresponding to each question, further extracting answer keywords from the obtained segmentation, and obtaining answer keywords corresponding to each question extracted by the target AI answering port reply client;
d5, matching the answer key words corresponding to the questions proposed by the target AI answering port answering client with the question key words corresponding to the questions, if the answer key words corresponding to the questions proposed by the target AI answering port answering client are successfully matched with the question key words corresponding to the questions, the answer correlation index corresponding to the questions is recorded as epsilon, otherwise, the answer correlation index corresponding to the questions is recorded as epsilon ', wherein epsilon is larger than epsilon';
d6, according to the answer correlation index corresponding to each question proposed by the target AI answering port answer client, counting the answer correlation degree of the question proposed by the target AI answering port answer client
Xi represents the degree of call-back association, λ, for the target AI listening port to answer the customer's proposed question
kThe answer correlation index, lambda, corresponding to the kth question is provided for the answer client of the target AI answering port
kThe value of (d) may be epsilon or epsilon'.
In the embodiment, the collection of the question and answer correlation degree of the reply client corresponding to the target AI answering port intuitively reflects the degree of question and answer correlation degree of the reply client in the answering call record of the target AI answering port, and provides the correlation coefficient of the answer correlation for the evaluation of the comprehensive answering service quality coefficient.
In this embodiment, the matching degree of the return voice parameters, the timeliness of the return call and the correlation degree of the return call are collected from the whole answering call record of the current calling client answered by the target AI answering port as the objective parameters of the AI answering the return call, because the matching degree of the return voice parameters can reflect the experience of the client on the speed and the tone of the return call of the target AI answering port, the greater the matching degree is, the better the client experiences the return voice of the target AI answering port, the timeliness of the return call can reflect the timeliness of the return call of the target AI answering port answering the problem posed by the client, the problem that the client is worried by waiting for a long time after the problem is posed, the call mood of the client is affected, the greater the timeliness is, the better the call mood of the client is, the correlation degree of the return call can reflect the correlation of the return call of the target AI answering port answering the problem posed by the client, and the correlation degree is greater, the more accurate the reply can be obtained by the client, so that the matching degree of the call back voice parameters, the call back timeliness and the call back correlation degree all have objective influence on the answering service quality output of the target AI answering port.
And the call ending service score acquisition module is used for acquiring the answering service score of the client on the target AI answering port after the answering call is ended.
The management database is used for storing the suitable answering speech parameters of the AI answering port corresponding to various speaking speech parameters of the customer, storing the standard comprehensive answering service quality coefficient, storing the evaluation weight proportion coefficient of the answering speech parameter matching degree, the answering timeliness, the answering relevance degree and the answering service score to the comprehensive answering service quality coefficient, and storing the bad ratio coefficient set value of the answering service.
The modeling processing module is used for evaluating the comprehensive answering service quality coefficient of the target AI answering port to the current calling client by combining the answering voice parameter matching degree corresponding to the target AI answering port, the answering timeliness of the target AI answering port for answering the question posed by the client, the answering relevance of the target AI answering port for answering the question posed by the client and the answering service score of the client to the target AI answering port
Expressed as the comprehensive answering service quality coefficient of the target AI answering port to the current calling client, q is expressed as the answering service grade value of the client to the target AI answering port
0Expressed as the highest score value of the answering service, alpha 1, alpha 2, alpha 03 and alpha 14 are expressed as the evaluation weight proportion coefficients of the answering voice parameter matching degree, the answering timeliness, the answering relevance degree and the answering service score to the comprehensive answering service quality coefficient, and alpha 21+ alpha 2+ alpha 3+ alpha 4 is 1, and the corresponding size relations of alpha 1, alpha 2, alpha 3 and alpha 4 are alpha 3>α2>α4>α1。
The embodiment determines a target AI answering port corresponding to a current calling client when a client call is received, extracts the whole answering call record of the target AI answering port for answering the current calling client, further processes the extracted answering call record, collects the matching degree of the answering voice parameters corresponding to the target AI answering port, the answering timeliness of the answering voice of the target AI answering port for answering the problem posed by the client and the answering relevance of the answering service of the target AI answering port for answering the problem posed by the client, and collects the answering service score of the client for the target AI answering port after the answering call is finished, thereby comprehensively evaluating the comprehensive answering service quality coefficient of the target AI answering port for the current calling client, realizing the comprehensive evaluation of the AI answering service quality according to the combination of the subjective answering service score and the AI answering objective answering call parameter of the client, the evaluation sidedness caused by the evaluation of the existing evaluation mode of the answering service quality of the AI call center only according to the subjective answering service score of the customer is avoided, so that the reliability of the evaluation result is improved.
The management server is used for numbering all AI answering ports in the AI calling center, counting the total times of answering client calling corresponding to all AI answering ports in a set time period, and further evaluating the comprehensive answering service quality coefficient corresponding to all the answering client calling in the set time period by all the AI answering ports according to the evaluation method of the comprehensive answering service quality coefficient of the target AI answering port to the current calling client, so as to identify the AI answering ports with poor answering service quality in the set time period according to the evaluation result, and the specific identification process comprises the following steps:
f1, comparing the comprehensive answering service quality coefficient corresponding to each answering client call in the set time period of each AI answering port with the set standard comprehensive answering service quality coefficient, and counting the times of answering client calls in the set time period of each AI answering port which are smaller than the set standard comprehensive answering service quality coefficient;
f2, dividing the total times of customer calls answered by each AI answering port in a set time period by the times of customer calls answered by the AI answering port in the set time period which is less than the set standard comprehensive answering service quality coefficient to obtain the corresponding answering service bad ratio coefficient of each AI answering port in the set time period;
f3, comparing the poor ratio coefficient of the answering service corresponding to each AI answering port in the set time period with the set value, if the poor ratio coefficient of the answering service corresponding to a certain AI answering port in the set time period is larger than the set value, the AI answering port is marked as the poor quality AI answering port.
In the embodiment, the comprehensive answering service quality coefficient corresponding to each time of answering a client call in a set time period by each AI answering port in the AI calling center is obtained by adopting the method for evaluating the comprehensive answering service quality coefficient of the target AI answering port to the current calling client, and then the AI answering port with poor answering service quality in the set time period is identified according to the evaluation result.
The display terminal is used for displaying the number corresponding to the AI answering port with poor answering service quality.
The foregoing is merely exemplary and illustrative of the present invention and various modifications, additions and substitutions may be made by those skilled in the art to the specific embodiments described without departing from the scope of the invention as defined in the following claims.