CN111292103A - Method and device for recommending customer service channel to user - Google Patents
Method and device for recommending customer service channel to user Download PDFInfo
- Publication number
- CN111292103A CN111292103A CN202010329615.4A CN202010329615A CN111292103A CN 111292103 A CN111292103 A CN 111292103A CN 202010329615 A CN202010329615 A CN 202010329615A CN 111292103 A CN111292103 A CN 111292103A
- Authority
- CN
- China
- Prior art keywords
- user
- channel
- self
- service channel
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06311—Scheduling, planning or task assignment for a person or group
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Educational Administration (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the specification provides a method and a device for recommending a customer service channel to a user, wherein the method comprises the following steps: acquiring a current user problem of a target user; determining the idle degree of the current artificial channel so as to obtain a first state characteristic; evaluating the idle degree of a future artificial channel so as to obtain a second state characteristic; determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user can receive a self-service channel when the self-service channel is recommended to the target user; determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel; and recommending a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user. The user can obtain good service experience.
Description
Technical Field
One or more embodiments of the present specification relate to the field of computers, and more particularly, to a method and apparatus for recommending a customer service channel to a user.
Background
Currently, with the expansion of business scale, more and more users dial a customer service hot line, and the proposed problems tend to be diversified. If all the questions of the user are answered by manual customer service, congestion often occurs during the customer service peak period, and therefore user experience is affected. With the continuous development of artificial intelligence technology, when users queue at customer service peak, the problem can be solved by partially adopting a mode of replacing an artificial channel with a self-service channel. In order to obtain good service experience, how to recommend a customer service channel to a user is a problem to be solved urgently.
Disclosure of Invention
One or more embodiments of the present specification describe a method and an apparatus for recommending a customer service channel to a user, which enable the user to obtain a good service experience.
In a first aspect, a method for recommending a customer service channel to a user is provided, where the customer service channel includes a self-service channel and a manual channel, and the method includes:
acquiring a current user problem of a target user;
determining the idle degree of the current artificial channel so as to obtain a first state characteristic;
evaluating the idle degree of a future artificial channel so as to obtain a second state characteristic;
determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user can accept a self-service channel when the self-service channel is recommended to the target user;
determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel;
recommending a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user.
In a possible implementation manner, the determining the idleness of the current artificial channel so as to obtain the first status characteristic includes:
acquiring at least one statistical index of the call completing rate of the current artificial channel, the number of people with idle artificial customer service equipped in the current artificial channel and the number of users waiting for recommending the customer service channel;
and determining the first state characteristic according to the at least one statistical index.
In a possible implementation manner, the evaluating the idleness of the future artificial channel to obtain the second state characteristic includes:
determining the number of first users waiting for recommending a customer service channel in the future by utilizing a pre-trained inflow prediction model according to the number of users waiting for recommending the customer service channel at present;
determining the number of second users to be born by the future artificial channel by utilizing a pre-trained carrying capacity prediction model according to the number of users to be born by the current artificial channel;
and determining the second state characteristic according to the number of the first users and the number of the second users.
Further, the inflow prediction model is obtained by training based on first training data; the first training data comprises historical change data of the number of users waiting for recommending the customer service channel.
Further, the carrying capacity prediction model is obtained by training based on second training data; the second training data comprises historical change data of the number of the users accepted by the artificial channel.
In one possible embodiment, the determining a corresponding third state feature according to the user representation data of the target user and the current user question includes:
acquiring user portrait data of the target user, a problem category corresponding to the current user problem and a user intention corresponding to the current user problem;
determining a first prediction result by utilizing a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category and the user intention, wherein the first prediction result is used for indicating whether the target user can accept a self-service channel when the target user recommends the self-service channel;
and determining the third state characteristic according to the first prediction result.
In a possible implementation, the third status feature is further used for indicating whether a self-service channel can solve the current user problem when the target user accepts the self-service channel;
determining a corresponding third state feature according to the user portrait data of the target user and the current user question, including:
acquiring user portrait data of the target user, a problem category corresponding to the current user problem and a user intention corresponding to the current user problem;
determining a first prediction result by utilizing a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category and the user intention, wherein the first prediction result is used for indicating whether the target user can accept a self-service channel when the target user recommends the self-service channel;
according to the user portrait data, the problem category and the user intention, determining a second prediction result by using a pre-trained self-help solution degree prediction model, wherein the second prediction result is used for indicating whether a self-help channel can solve the current user problem or not when the target user accepts the self-help channel;
and determining the third state characteristic according to the first prediction result and the second prediction result.
In a possible implementation manner, after recommending a self-service channel or a manual channel to the target user, the method further includes:
determining rewards corresponding to the states and the target actions according to whether the call completing rate of the current artificial channel belongs to a preset interval or not;
and training the reinforcement learning model according to the reward.
In a second aspect, an apparatus for recommending a customer service channel to a user is provided, where the customer service channel includes a self-service channel and a manual channel, and the apparatus includes:
the acquisition unit is used for acquiring the current user problem of the target user;
the first feature extraction unit is used for determining the idle degree of the current artificial channel so as to obtain a first state feature;
the second feature extraction unit is used for evaluating the idle degree of the future artificial channel so as to obtain a second state feature;
a third feature extraction unit, configured to determine a corresponding third state feature according to the user portrait data of the target user and the current user problem acquired by the acquisition unit, where the third state feature is used to indicate whether the target user will accept a self-service channel when the self-service channel is recommended to the target user;
the determining unit is used for determining the state of a reinforcement learning model according to the first state feature obtained by the first feature extracting unit, the second state feature obtained by the second feature extracting unit and the third state feature obtained by the third feature extracting unit, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and an artificial channel;
and the recommending unit is used for recommending a self-service channel or a manual channel to the target user according to the target action obtained by the determining unit so as to solve the problem of the current user of the target user.
In a third aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of the first aspect.
In a fourth aspect, there is provided a computing device comprising a memory having stored therein executable code and a processor that, when executing the executable code, implements the method of the first aspect.
According to the method and the device provided by the embodiment of the specification, the current user problem of the target user is firstly obtained; then determining the idle degree of the current artificial channel so as to obtain a first state characteristic; evaluating the idle degree of the future artificial channel so as to obtain a second state characteristic; determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user can accept a self-service channel when the self-service channel is recommended to the target user; then, determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel; and finally recommending a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user. As can be seen from the above, in the embodiment of the present specification, when a current user problem of a target user is faced, a self-service channel or a manual channel is selected to be recommended to the target user through a reinforcement learning model, so that an accurate control purpose can be achieved. The state of the reinforcement learning model is determined based on the first state feature, the second state feature and the third state feature, wherein the first state feature and the second state feature respectively consider the idle degree of the current artificial channel and the idle degree of the future artificial channel, so that the call completing rate of the artificial channel is favorably controlled in a reasonable interval range, and a user can obtain good service experience. In addition, the acceptance of the target user to the self-service channel is considered by the third state characteristic, and the user experience is further improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic diagram illustrating an implementation scenario of an embodiment disclosed herein;
FIG. 2 illustrates a flow diagram of a method for recommending a customer service channel to a user, according to one embodiment;
FIG. 3 illustrates a schematic diagram of a manner of determining a first status feature, according to one embodiment;
FIG. 4 illustrates a schematic diagram of a manner of determining a second status feature, according to one embodiment;
FIG. 5 illustrates a schematic diagram of a determination of a third status feature according to one embodiment;
FIG. 6 shows a schematic block diagram of an apparatus for recommending a customer service channel to a user according to one embodiment.
Detailed Description
The scheme provided by the specification is described below with reference to the accompanying drawings.
Fig. 1 is a schematic view of an implementation scenario of an embodiment disclosed in this specification. The implementation scenario involves recommending a customer service channel to a user. In the embodiment of the specification, the customer service channel comprises a self-service channel and a manual channel. It can be understood that the manual channel is just that the manual customer service solves the problem for the user, while the self-service channel is not that the manual customer service solves the problem for the user, and the self-service channel is that the machine replaces the manual customer service to solve the problem for the user, and can include the following forms: the robot is used for answering with the human to solve the problem for the user; or the machine sends a short message containing a link to the user, and the user acquires an answer to the question by clicking the link; alternatively, the user interacts with the machine through a question and answer page to obtain answers to the questions.
Referring to fig. 1, after a current user problem of a target user is obtained, a first state feature corresponding to a current field state, a second state feature corresponding to a predicted field state, and a third state feature corresponding to a user state are obtained, a state of a reinforcement learning model is determined according to the first state feature, the second state feature, and the third state feature, and a model decision is made through the reinforcement learning model, that is, a self-service channel or an artificial channel is recommended to the target user, so as to solve the current user problem of the target user.
In the embodiment of the specification, considering that for different users, different problems, different time and the like, the users queue up abandoned artificial channels, the accepting willingness of the self-service channels is different, and even after the users accept the self-service channels, the degree of whether the users can really help the users to solve the problems is different, so that whether the self-service channels are recommended to the users is determined according to the problems of the target users and the current users, and the problem that the self-service channels are recommended to the target users when the target users are not suitable for the self-service channels or the current users are not suitable for the self-service channels is avoided, the time of the users is delayed when the self-service.
In addition, the call completing rate is usually used as an index for measuring customer service congestion, and if the call completing rate is too high (such as close to 100%), it indicates that the field is too idle, and resources of artificial customer service are not fully utilized; if the call completing rate is too low, the situation is indicated to be too congested, and the waiting time of the user is too long. Therefore, the embodiment of the specification is equivalent to that a control closed loop controls the call completing rate within a reasonable and healthy interval range by recommending a self-service channel or a manual channel to the target user. The closed control loop realizes the balance of the call completing rate by distributing proper users to self-service channels and manual channels according to the current call completing rate.
Reinforcement Learning (RL) is one of machine learning methods, and is mainly characterized in that an agent learns in a trial and error manner, and a reward guidance behavior obtained by interacting with the environment aims to enable the agent to obtain the maximum reward.
Closed loop control refers to a basic concept of control theory. A control relationship in which an output as a controlled is returned to an input as a control in a certain manner and a control influence is exerted on the input, and a system control mode with feedback information. When the operator starts the system, the control information is transmitted to the controlled object through the system operation, and the state information of the controlled object is fed back to the input to correct the operation process, so that the output of the system meets the expected requirement.
Fig. 2 is a flowchart illustrating a method for recommending a customer service channel to a user, wherein the customer service channel includes a self-service channel and a manual channel, and the method can be based on the implementation scenario illustrated in fig. 1. As shown in fig. 2, the method for recommending a customer service channel to a user in this embodiment includes the following steps: step 21, obtaining the current user problem of the target user; step 22, determining the idle degree of the current artificial channel so as to obtain a first state characteristic; step 23, evaluating the idle degree of the future artificial channel so as to obtain a second state characteristic; step 24, determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user can accept a self-service channel when the self-service channel is recommended to the target user; step 25, determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel; and step 26, recommending a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user. Specific execution modes of the above steps are described below.
First, in step 21, the current user question of the target user is obtained. It is understood that, before recommending a customer service channel for a target user, a preliminary description of a current user problem of the target user may be obtained, for example, a robot may communicate with the target user to guide the target user to make the preliminary description of the current user problem.
As an example, the above-mentioned preliminary description of the current user question may embody a question category corresponding to the current user question, which may correspond to a large business category, for example, the question category is insurance business; and/or the preliminary description of the current user question may reflect a user intention corresponding to the current user question, where the user intention may be more precise, for example, the user intention is a standard question, and the user intention may also be more extensive, for example, the user intention is a business element or a appeal element.
It will be appreciated that standard question sentences are questions that some users may ask organized by business. The service element refers to a service category of the branch service corresponding to the standard question, for example, mutual insurance, and the appeal element refers to a requirement or intention of the user, such as rebate.
Then, in step 22, the idleness of the current artificial channel is determined, so as to obtain a first status characteristic. It is understood that the vacancy degree of the current artificial channel corresponds to the congestion degree of the current artificial channel, and the first status feature is used for representing the vacancy degree of the current artificial channel, that is, the first status feature is used for representing the congestion degree of the current artificial channel.
In one example, the determining the idleness of the current artificial channel to obtain the first status feature comprises:
acquiring at least one statistical index of the call completing rate of the current artificial channel, the number of people with idle artificial customer service equipped in the current artificial channel and the number of users waiting for recommending the customer service channel;
and determining the first state characteristic according to the at least one statistical index.
This example may determine the first state characteristic via one or more statistical indicators. Optionally, the first state characteristic is determined by integrating a plurality of statistical indexes, so that the first state characteristic can accurately represent the idle degree of the current artificial channel.
FIG. 3 illustrates a schematic diagram of a determination of a first status feature according to one embodiment. Referring to fig. 3, a multidimensional vector is determined according to the call completing rate of the current artificial channel, the number of people with artificial customer service idles equipped in the current artificial channel, and the number of users waiting for the recommended customer service channel, and the multidimensional vector is used as the first state feature. It can be understood that the call completing rate of the current artificial channel, the number of people with artificial customer service being available in the current artificial channel, and the number of users waiting for the recommended customer service channel respectively correspond to one or more dimensions of the multi-dimensional vector.
Next, in step 23, the idleness of the future artificial channel is evaluated, so as to obtain a second status feature. It is understood that the idle degree of the future artificial channel is the idle degree of the artificial channel after a preset time period, and the preset time period may be set according to the user requirement, for example, set to 1 minute, 2 minutes, or 5 minutes, etc.
In one example, the evaluating the idleness of the future artificial channel to obtain the second status feature comprises:
determining the number of first users waiting for recommending a customer service channel in the future by utilizing a pre-trained inflow prediction model according to the number of users waiting for recommending the customer service channel at present;
determining the number of second users to be born by the future artificial channel by utilizing a pre-trained carrying capacity prediction model according to the number of users to be born by the current artificial channel;
and determining the second state characteristic according to the number of the first users and the number of the second users.
Further, the inflow prediction model is obtained by training based on first training data; the first training data comprises historical change data of the number of users waiting for recommending the customer service channel.
Further, the carrying capacity prediction model is obtained by training based on second training data; the second training data comprises historical change data of the number of the users accepted by the artificial channel.
FIG. 4 illustrates a schematic diagram of a determination of a second status feature according to one embodiment. Referring to fig. 4, according to the current inflow and the current receiving capacity, a first prediction model is used to predict a future inflow and a future receiving capacity, wherein the first prediction model is equivalent to integrate the inflow prediction model and the receiving capacity prediction model into a single model, the inflow is the number of first users waiting for the recommended customer service channel, the receiving capacity is the number of second users receiving the artificial channel, and the predicted future inflow minus the predicted future receiving capacity indicates a future receiving gap. In this embodiment, the second status feature may correspond to a specific numerical value, where the numerical value is used to indicate the receiving gap, or the second status feature may correspond to a multidimensional vector, where the number of the first user and the number of the second user respectively correspond to one or more dimensions of the multidimensional vector.
And step 24, determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user receives a self-service channel when the self-service channel is recommended to the target user. It is to be appreciated that after recommending a self-service channel to a user, the user may either agree to employ the self-service channel or refuse to employ the self-service channel.
In one example, the determining a corresponding third state feature from the user representation data of the target user and the current user question comprises:
acquiring user portrait data of the target user, a problem category corresponding to the current user problem and a user intention corresponding to the current user problem;
determining a first prediction result by utilizing a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category and the user intention, wherein the first prediction result is used for indicating whether the target user can accept a self-service channel when the target user recommends the self-service channel;
and determining the third state characteristic according to the first prediction result.
It will be appreciated that the self-help receptivity prediction model described above may be trained using training data comprising: historical user portrait data, problem categories and user intentions are used as sample characteristics, whether a user historically accepts a self-service channel is used as a label, the label of the user historically accepting the self-service channel is 1, and the label of the user historically not accepting the self-service channel is 0.
By way of example, the user representation data may include the user's age, gender, or occupation, etc.; the question categories and user intentions may be identified by a neural network model.
In another example, the third status feature is further used to indicate whether a self-service channel can solve the current user problem when the target user accepts the self-service channel;
determining a corresponding third state feature according to the user portrait data of the target user and the current user question, including:
acquiring user portrait data of the target user, a problem category corresponding to the current user problem and a user intention corresponding to the current user problem;
determining a first prediction result by utilizing a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category and the user intention, wherein the first prediction result is used for indicating whether the target user can accept a self-service channel when the target user recommends the self-service channel;
according to the user portrait data, the problem category and the user intention, determining a second prediction result by using a pre-trained self-help solution degree prediction model, wherein the second prediction result is used for indicating whether a self-help channel can solve the current user problem or not when the target user accepts the self-help channel;
and determining the third state characteristic according to the first prediction result and the second prediction result.
It will be appreciated that the self-help solution prediction model described above may be trained using training data including: historical user portrait data, problem categories and user intentions are used as sample characteristics, whether a user problem is solved or not is used as a label in the user history under a self-service channel, the label of the user problem is solved as 1 in the user history under the self-service channel, and the label of the user problem which is not solved in the user history under the self-service channel is 0.
FIG. 5 illustrates a schematic diagram of a determination of a third status feature according to one embodiment. Referring to fig. 5, according to the user portrait data, the problem category and the user intention, a self-help acceptance and a self-help solution are respectively predicted by using a second prediction model, wherein the second prediction model is equivalent to integrating the self-help acceptance prediction model and the self-help solution prediction model into one model, the self-help acceptance is the first prediction result, and the self-help solution is the second prediction result, and the self-help acceptance and the self-help solution may be in a probability form. In this embodiment, the third state feature may correspond to a multidimensional vector, and the first prediction result and the second prediction result respectively correspond to one or more dimensions of the multidimensional vector.
And step 25, determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel. It is understood that the process of the reinforcement learning model outputting the target action is the process of making the model decision.
In the embodiment of the present specification, the reinforcement learning model may output different target actions corresponding to different states, and thus, different target users may output different target actions; different current user problems of the same target user may output different target actions; the same target user may output different target actions at different times than the current user question.
Finally, at step 26, a self-service channel or a manual channel is recommended to the target user according to the target action, so as to solve the current user problem of the target user. It is to be appreciated that after recommending a self-service channel to a user, the user may either agree to employ the self-service channel or refuse to employ the self-service channel.
In one example, if a user declines to employ a self-service channel, a manual channel is employed to provide customer service to the user.
In one example, after recommending a self-service channel or a manual channel to the target user, the method further comprises:
determining rewards corresponding to the states and the target actions according to whether the call completing rate of the current artificial channel belongs to a preset interval or not;
and training the reinforcement learning model according to the reward.
It is to be understood that the call completing rate may be determined periodically, and the reinforcement learning model may be trained periodically.
According to the method provided by the embodiment of the specification, the current user problem of the target user is obtained firstly; then determining the idle degree of the current artificial channel so as to obtain a first state characteristic; evaluating the idle degree of the future artificial channel so as to obtain a second state characteristic; determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user can accept a self-service channel when the self-service channel is recommended to the target user; then, determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel; and finally recommending a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user. As can be seen from the above, in the embodiment of the present specification, when a current user problem of a target user is faced, a self-service channel or a manual channel is selected to be recommended to the target user through a reinforcement learning model, so that an accurate control purpose can be achieved. The state of the reinforcement learning model is determined based on the first state feature, the second state feature and the third state feature, wherein the first state feature and the second state feature respectively consider the idle degree of the current artificial channel and the idle degree of the future artificial channel, so that the call completing rate of the artificial channel is favorably controlled in a reasonable interval range, and a user can obtain good service experience. In addition, the acceptance of the target user to the self-service channel is considered by the third state characteristic, and the user experience is further improved.
According to an embodiment of another aspect, an apparatus for recommending a customer service channel to a user is also provided, and the apparatus is configured to execute the method for recommending a customer service channel to a user provided in the embodiment of the present specification. FIG. 6 shows a schematic block diagram of an apparatus for recommending a customer service channel to a user according to one embodiment. As shown in fig. 6, the apparatus 600 includes:
an obtaining unit 61, configured to obtain a current user question of a target user;
the first feature extraction unit 62 is configured to determine an idle degree of the current artificial channel, so as to obtain a first state feature;
the second feature extraction unit 63 is configured to evaluate an idle degree of a future artificial channel, so as to obtain a second state feature;
a third feature extraction unit 64, configured to determine a corresponding third status feature according to the user portrait data of the target user and the current user problem acquired by the acquisition unit 61, where the third status feature is used to indicate whether the target user will accept a self-service channel when the self-service channel is recommended to the target user;
a determining unit 65, configured to determine a state of a reinforcement learning model according to the first state feature obtained by the first feature extracting unit 62, the second state feature obtained by the second feature extracting unit 63, and the third state feature obtained by the third feature extracting unit 64, and output a target action in a selectable action set through the reinforcement learning model, where the selectable action set includes a self-service channel and an artificial channel;
a recommending unit 66, configured to recommend a self-service channel or a manual channel to the target user according to the target action obtained by the determining unit 65, so as to solve the current user problem of the target user.
Optionally, as an embodiment, the first feature extraction unit 62 specifically includes:
the index acquisition subunit is used for acquiring at least one statistical index of the call completing rate of the current artificial channel, the number of people with idle artificial customer service equipped in the current artificial channel and the number of users waiting for the recommended customer service channel;
the first determining subunit is configured to determine the first state feature according to the at least one statistical indicator acquired by the indicator acquiring subunit.
Optionally, as an embodiment, the second feature extraction unit 63 specifically includes:
the inflow prediction subunit is used for determining the number of first users waiting for the recommended customer service channel in the future by utilizing a pre-trained inflow prediction model according to the number of users waiting for the recommended customer service channel currently;
the receiving capacity prediction subunit is used for determining the number of second users to be received by the future artificial channel by utilizing a pre-trained receiving capacity prediction model according to the number of users to be received by the current artificial channel;
and the second determining subunit is used for determining the second state characteristic according to the first user number determined by the inflow prediction subunit and the second user number determined by the carrying capacity prediction subunit.
Further, the inflow prediction model is obtained by training based on first training data; the first training data comprises historical change data of the number of users waiting for recommending the customer service channel.
Further, the carrying capacity prediction model is obtained by training based on second training data; the second training data comprises historical change data of the number of the users accepted by the artificial channel.
Optionally, as an embodiment, the third feature extraction unit 64 specifically includes:
a user data obtaining subunit, configured to obtain user portrait data of the target user, a question category corresponding to the current user question, and a user intention corresponding to the current user question;
the acceptance prediction subunit is configured to determine a first prediction result by using a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category, and the user intention acquired by the user data acquisition subunit, where the first prediction result is used to indicate whether the target user will accept a self-service channel when recommending the self-service channel to the target user;
and the third determining subunit is configured to determine the third state characteristic according to the first prediction result obtained by the receptivity predicting subunit.
Optionally, as an embodiment, the third status feature is further used to indicate whether a self-service channel can solve the current user problem when the target user accepts the self-service channel;
the third feature extraction unit 64 specifically includes:
a user data obtaining subunit, configured to obtain user portrait data of the target user, a question category corresponding to the current user question, and a user intention corresponding to the current user question;
the acceptance prediction subunit is configured to determine a first prediction result by using a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category, and the user intention acquired by the user data acquisition subunit, where the first prediction result is used to indicate whether the target user will accept a self-service channel when recommending the self-service channel to the target user;
a resolution prediction subunit, configured to determine, according to the user portrait data, the problem category, and the user intention acquired by the user data acquisition subunit, a second prediction result by using a pre-trained self-service resolution prediction model, where the second prediction result is used to indicate whether a self-service channel can solve the current user problem when the target user accepts the self-service channel;
and a fourth determining subunit that determines the third state feature according to the first prediction result obtained by the receptivity prediction subunit and the second prediction result obtained by the resolution prediction subunit.
Optionally, as an embodiment, the apparatus further includes:
a reward determining unit, configured to determine, after the recommending unit 66 recommends a self-service channel or an artificial channel to the target user, a reward corresponding to the state and the target action according to whether a call completing rate of a current artificial channel belongs to a preset interval;
and the training unit is used for training the reinforcement learning model according to the reward determined by the reward determining unit.
With the apparatus provided in the embodiment of the present specification, first, the obtaining unit 61 obtains a current user question of a target user; then the first feature extraction unit 62 determines the idle degree of the current artificial channel, so as to obtain a first state feature; the second feature extraction unit 63 evaluates the idle degree of the future artificial channel to obtain a second state feature; the third feature extraction unit 64 determines a corresponding third state feature according to the user portrait data of the target user and the current user problem, wherein the third state feature is used for indicating whether the target user can accept a self-service channel when the self-service channel is recommended to the target user; then the determining unit 65 determines the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputs a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel; and finally, the recommending unit 66 recommends a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user. As can be seen from the above, in the embodiment of the present specification, when a current user problem of a target user is faced, a self-service channel or a manual channel is selected to be recommended to the target user through a reinforcement learning model, so that an accurate control purpose can be achieved. The state of the reinforcement learning model is determined based on the first state feature, the second state feature and the third state feature, wherein the first state feature and the second state feature respectively consider the idle degree of the current artificial channel and the idle degree of the future artificial channel, so that the call completing rate of the artificial channel is favorably controlled in a reasonable interval range, and a user can obtain good service experience. In addition, the acceptance of the target user to the self-service channel is considered by the third state characteristic, and the user experience is further improved.
According to an embodiment of another aspect, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method described in connection with fig. 2.
According to an embodiment of yet another aspect, there is also provided a computing device comprising a memory having stored therein executable code, and a processor that, when executing the executable code, implements the method described in connection with fig. 2.
Those skilled in the art will recognize that, in one or more of the examples described above, the functions described in this invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present invention should be included in the scope of the present invention.
Claims (18)
1. A method of recommending a customer service channel to a user, the customer service channel comprising a self-service channel and a manual channel, the method comprising:
acquiring a current user problem of a target user;
determining the idle degree of the current artificial channel so as to obtain a first state characteristic;
evaluating the idle degree of a future artificial channel so as to obtain a second state characteristic;
determining a corresponding third state characteristic according to the user portrait data of the target user and the current user problem, wherein the third state characteristic is used for indicating whether the target user can accept a self-service channel when the self-service channel is recommended to the target user;
determining the state of a reinforcement learning model according to the first state feature, the second state feature and the third state feature, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and a manual channel;
recommending a self-service channel or a manual channel to the target user according to the target action so as to solve the current user problem of the target user.
2. The method of claim 1, wherein the determining the idleness of the current artificial channel to obtain the first status characteristic comprises:
acquiring at least one statistical index of the call completing rate of the current artificial channel, the number of people with idle artificial customer service equipped in the current artificial channel and the number of users waiting for recommending the customer service channel;
and determining the first state characteristic according to the at least one statistical index.
3. The method of claim 1, wherein said evaluating the idleness of the future artificial channel to derive the second status characteristic comprises:
determining the number of first users waiting for recommending a customer service channel in the future by utilizing a pre-trained inflow prediction model according to the number of users waiting for recommending the customer service channel at present;
determining the number of second users to be born by the future artificial channel by utilizing a pre-trained carrying capacity prediction model according to the number of users to be born by the current artificial channel;
and determining the second state characteristic according to the number of the first users and the number of the second users.
4. The method of claim 3, wherein the inflow prediction model is trained based on first training data; the first training data comprises historical change data of the number of users waiting for recommending the customer service channel.
5. The method of claim 3, wherein the bearing prediction model is trained based on second training data; the second training data comprises historical change data of the number of the users accepted by the artificial channel.
6. The method of claim 1, wherein said determining a corresponding third state feature from the user representation data of the target user and the current user question comprises:
acquiring user portrait data of the target user, a problem category corresponding to the current user problem and a user intention corresponding to the current user problem;
determining a first prediction result by utilizing a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category and the user intention, wherein the first prediction result is used for indicating whether the target user can accept a self-service channel when the target user recommends the self-service channel;
and determining the third state characteristic according to the first prediction result.
7. The method of claim 1, wherein the third status feature is further used to indicate whether a self-service channel can resolve the current user problem when the target user accepts the self-service channel;
determining a corresponding third state feature according to the user portrait data of the target user and the current user question, including:
acquiring user portrait data of the target user, a problem category corresponding to the current user problem and a user intention corresponding to the current user problem;
determining a first prediction result by utilizing a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category and the user intention, wherein the first prediction result is used for indicating whether the target user can accept a self-service channel when the target user recommends the self-service channel;
according to the user portrait data, the problem category and the user intention, determining a second prediction result by using a pre-trained self-help solution degree prediction model, wherein the second prediction result is used for indicating whether a self-help channel can solve the current user problem or not when the target user accepts the self-help channel;
and determining the third state characteristic according to the first prediction result and the second prediction result.
8. The method of claim 1, wherein after the recommending a self-service channel or a human channel to the target user, the method further comprises:
determining rewards corresponding to the states and the target actions according to whether the call completing rate of the current artificial channel belongs to a preset interval or not;
and training the reinforcement learning model according to the reward.
9. An apparatus for recommending a customer service channel to a user, the customer service channel comprising a self-service channel and a manual channel, the apparatus comprising:
the acquisition unit is used for acquiring the current user problem of the target user;
the first feature extraction unit is used for determining the idle degree of the current artificial channel so as to obtain a first state feature;
the second feature extraction unit is used for evaluating the idle degree of the future artificial channel so as to obtain a second state feature;
a third feature extraction unit, configured to determine a corresponding third state feature according to the user portrait data of the target user and the current user problem acquired by the acquisition unit, where the third state feature is used to indicate whether the target user will accept a self-service channel when the self-service channel is recommended to the target user;
the determining unit is used for determining the state of a reinforcement learning model according to the first state feature obtained by the first feature extracting unit, the second state feature obtained by the second feature extracting unit and the third state feature obtained by the third feature extracting unit, and outputting a target action in a selectable action set through the reinforcement learning model, wherein the selectable action set comprises a self-service channel and an artificial channel;
and the recommending unit is used for recommending a self-service channel or a manual channel to the target user according to the target action obtained by the determining unit so as to solve the problem of the current user of the target user.
10. The apparatus according to claim 9, wherein the first feature extraction unit specifically includes:
the index acquisition subunit is used for acquiring at least one statistical index of the call completing rate of the current artificial channel, the number of people with idle artificial customer service equipped in the current artificial channel and the number of users waiting for the recommended customer service channel;
the first determining subunit is configured to determine the first state feature according to the at least one statistical indicator acquired by the indicator acquiring subunit.
11. The apparatus according to claim 9, wherein the second feature extraction unit specifically includes:
the inflow prediction subunit is used for determining the number of first users waiting for the recommended customer service channel in the future by utilizing a pre-trained inflow prediction model according to the number of users waiting for the recommended customer service channel currently;
the receiving capacity prediction subunit is used for determining the number of second users to be received by the future artificial channel by utilizing a pre-trained receiving capacity prediction model according to the number of users to be received by the current artificial channel;
and the second determining subunit is used for determining the second state characteristic according to the first user number determined by the inflow prediction subunit and the second user number determined by the carrying capacity prediction subunit.
12. The apparatus of claim 11, wherein the inflow prediction model is trained based on first training data; the first training data comprises historical change data of the number of users waiting for recommending the customer service channel.
13. The apparatus of claim 11, wherein the bearing prediction model is trained based on second training data; the second training data comprises historical change data of the number of the users accepted by the artificial channel.
14. The apparatus according to claim 9, wherein the third feature extraction unit specifically includes:
a user data obtaining subunit, configured to obtain user portrait data of the target user, a question category corresponding to the current user question, and a user intention corresponding to the current user question;
the acceptance prediction subunit is configured to determine a first prediction result by using a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category, and the user intention acquired by the user data acquisition subunit, where the first prediction result is used to indicate whether the target user will accept a self-service channel when recommending the self-service channel to the target user;
and the third determining subunit is configured to determine the third state characteristic according to the first prediction result obtained by the receptivity predicting subunit.
15. The apparatus of claim 9, wherein the third status feature is further to indicate whether a self-service channel is able to resolve the current user issue when the target user accepts the self-service channel;
the third feature extraction unit specifically includes:
a user data obtaining subunit, configured to obtain user portrait data of the target user, a question category corresponding to the current user question, and a user intention corresponding to the current user question;
the acceptance prediction subunit is configured to determine a first prediction result by using a pre-trained self-service acceptance prediction model according to the user portrait data, the problem category, and the user intention acquired by the user data acquisition subunit, where the first prediction result is used to indicate whether the target user will accept a self-service channel when recommending the self-service channel to the target user;
a resolution prediction subunit, configured to determine, according to the user portrait data, the problem category, and the user intention acquired by the user data acquisition subunit, a second prediction result by using a pre-trained self-service resolution prediction model, where the second prediction result is used to indicate whether a self-service channel can solve the current user problem when the target user accepts the self-service channel;
and a fourth determining subunit that determines the third state feature according to the first prediction result obtained by the receptivity prediction subunit and the second prediction result obtained by the resolution prediction subunit.
16. The apparatus of claim 9, wherein the apparatus further comprises:
the reward determining unit is used for determining rewards corresponding to the state and the target action according to whether the call-in rate of the current artificial channel belongs to a preset interval or not after the recommending unit recommends the self-service channel or the artificial channel to the target user;
and the training unit is used for training the reinforcement learning model according to the reward determined by the reward determining unit.
17. A computer-readable storage medium, on which a computer program is stored which, when executed in a computer, causes the computer to carry out the method of any one of claims 1-8.
18. A computing device comprising a memory having stored therein executable code and a processor that, when executing the executable code, implements the method of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010329615.4A CN111292103A (en) | 2020-04-24 | 2020-04-24 | Method and device for recommending customer service channel to user |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010329615.4A CN111292103A (en) | 2020-04-24 | 2020-04-24 | Method and device for recommending customer service channel to user |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111292103A true CN111292103A (en) | 2020-06-16 |
Family
ID=71027409
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010329615.4A Pending CN111292103A (en) | 2020-04-24 | 2020-04-24 | Method and device for recommending customer service channel to user |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111292103A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112070518A (en) * | 2020-11-11 | 2020-12-11 | 支付宝(杭州)信息技术有限公司 | Customer service providing method and system |
| CN113077169A (en) * | 2021-04-16 | 2021-07-06 | 上海华客信息科技有限公司 | Method, system, equipment and medium for recommending customer service personnel based on label |
| CN117788628A (en) * | 2024-02-27 | 2024-03-29 | 厦门众联世纪股份有限公司 | Creative material generation method based on AIGC |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140355750A1 (en) * | 2013-05-28 | 2014-12-04 | Oracle International Corporation | Contact center skills modeling using customer relationship management (crm) incident categorization structure |
| CN107301213A (en) * | 2017-06-09 | 2017-10-27 | 腾讯科技（深圳）有限公司 | Intelligent answer method and device |
| CN109635271A (en) * | 2018-10-22 | 2019-04-16 | 阿里巴巴集团控股有限公司 | A kind of user's intension recognizing method, customer service system, device and electronic equipment |
-
2020
- 2020-04-24 CN CN202010329615.4A patent/CN111292103A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140355750A1 (en) * | 2013-05-28 | 2014-12-04 | Oracle International Corporation | Contact center skills modeling using customer relationship management (crm) incident categorization structure |
| CN107301213A (en) * | 2017-06-09 | 2017-10-27 | 腾讯科技（深圳）有限公司 | Intelligent answer method and device |
| CN109635271A (en) * | 2018-10-22 | 2019-04-16 | 阿里巴巴集团控股有限公司 | A kind of user's intension recognizing method, customer service system, device and electronic equipment |
Non-Patent Citations (2)
| Title |
|---|
| ZINING LIU 等: "Which Channel to Ask My Question?: Personalized Customer Service Request Stream Routing Using Deep Reinforcement Learning", 《IEEE ACCESS》 * |
| 王颖 等: "基于用户行为分析的服务渠道评价研究与实现", 《陕西电力》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112070518A (en) * | 2020-11-11 | 2020-12-11 | 支付宝(杭州)信息技术有限公司 | Customer service providing method and system |
| CN113077169A (en) * | 2021-04-16 | 2021-07-06 | 上海华客信息科技有限公司 | Method, system, equipment and medium for recommending customer service personnel based on label |
| CN117788628A (en) * | 2024-02-27 | 2024-03-29 | 厦门众联世纪股份有限公司 | Creative material generation method based on AIGC |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111292103A (en) | Method and device for recommending customer service channel to user | |
| US11770476B2 (en) | Intent analysis for call center response generation | |
| CN110322169B (en) | Task issuing method and device | |
| CN113031983B (en) | Intelligent software upgrading method and device based on deep reinforcement learning | |
| Alshahrani et al. | Validity, reliability, predictors, moderation: The UTAUT model revisited | |
| US20200104733A1 (en) | Generation of human readable explanations of data driven analytics | |
| CN111400466A (en) | Intelligent dialogue method and device based on reinforcement learning | |
| Kendall et al. | On the complexity of forming mental models | |
| CN111046156B (en) | Method, device and server for determining rewarding data | |
| Dummel et al. | Effects of ego-depletion on choice behaviour in a multi-attribute decision task | |
| CN112015861A (en) | Intelligent test paper algorithm based on user historical behavior analysis | |
| CN113537631B (en) | Medicine demand prediction method, device, electronic equipment and storage medium | |
| CN112734142B (en) | Resource learning path planning method and device based on deep learning | |
| CN114090733A (en) | Learning resource recommendation method and device, storage medium and electronic equipment | |
| CN111382263B (en) | Data processing method, device and system | |
| CN113868534A (en) | Service processing method, service processing device, electronic device and storage medium | |
| US20200401937A1 (en) | Information processing apparatus and non-transitory computer readable medium | |
| JP6934351B2 (en) | AI service usage support system | |
| Spicer et al. | Repeated risky choices become more consistent with themselves but not expected value, with no effect of matched trial order | |
| Lee et al. | Choice-induced preference change under a sequential sampling model framework | |
| CN115515088B (en) | Method for arranging business process chain, and method and device for interacting 5G messages | |
| US20240012841A1 (en) | Method and system for integrated monitoring of chatbots for concept drift detection | |
| CN109783627B (en) | Automatic response method, device, computer equipment and storage medium | |
| CN117786408A (en) | Training method and device for large language model, electronic equipment and storage medium | |
| US20230033737A1 (en) | Method and system of compliance scenario prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200616 |
|
| RJ01 | Rejection of invention patent application after publication |