CN116108172A

CN116108172A - Task management method, device, equipment and storage medium based on machine learning

Info

Publication number: CN116108172A
Application number: CN202211662559.1A
Authority: CN
Inventors: 戴耀康; 李子旺; 余浩然; 何沛钊; 杨飞彬; 陈剑生; 刘杭; 朱潇然; 高海滔; 刘锦健; 钟平彬; 肖冬晋; 陈厚因
Original assignee: Peoples Insurance Company of China
Current assignee: Peoples Insurance Company of China
Priority date: 2022-12-23
Filing date: 2022-12-23
Publication date: 2023-05-12

Abstract

The application discloses a task management method based on machine learning, which is used for solving the problem that the service efficiency is greatly reduced because the conventional task switching management scheme mainly depends on manual setting of a background manager and the task management scheduling mode is single. The method comprises the following steps: responding to a received first task, and initiating a voice call request to a user corresponding to the first task; according to a pre-constructed question library, interacting with the user to obtain question and answer data of the user in the interaction process; processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data; judging whether task allocation is carried out for the first task according to the user intention; and when the judgment result is yes, determining the artificial seat corresponding to the first task, and allocating the first task to the artificial seat for processing.

Description

Task management method, device, equipment and storage medium based on machine learning

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a task management method, device, equipment and storage medium based on machine learning.

Background

With the development of current artificial intelligence technologies such as automatic Speech recognition technology (Automatic Speech Recognition, ASR), text To Speech technology (TTS), and natural language processing technology (Natural Language Processing, NLP), direct voice communication between computers and people is becoming a reality. Based on these technologies, the developed intelligent voice interactive robots have replaced artificial agents to provide services for users in various scenarios such as sales, customer service, and call centers.

The intelligent voice interaction robot has the advantages that the intelligent voice interaction robot can automatically dial according to a pre-configured task, so that the calling efficiency is greatly improved, more clients are touched, and the labor cost is greatly reduced. In the prior art, the intelligent voice interactive robot is limited by the accuracy of voice recognition and semantic understanding of the robot and the content quantity of a knowledge base, and in some complex interaction scenes, when the intelligent voice interactive robot faces complex problems or unrecognizable intentions, the situation that users answer questions or cannot accurately communicate with clients can occur, and when users perceive the situations, the situation that communication will is reduced, the users feel opposite or even hang up calls directly can occur, so that the loss of the intended clients is caused.

Therefore, in the existing scheme, the intelligent voice interaction robot cannot be used for directly and completely replacing the artificial seat, or the switching between the intelligent voice interaction robot and the artificial seat is required according to actual needs.

However, at present, the task switching mode between the intelligent voice interaction robot and the artificial seat is single, and mainly depends on the manual setting of a background manager, so that effective bidirectional interaction communication between the intelligent voice interaction robot and the artificial seat is lacking, the accuracy, the instantaneity and the scientificity of the conventional task allocation strategy are lacking, and the service efficiency of the intelligent voice interaction robot and the artificial seat is greatly reduced.

Therefore, how to efficiently, accurately and dynamically complete task scheduling management between the intelligent voice interaction robot and the artificial seat in real time becomes a problem to be solved by related technicians in the field at present.

Disclosure of Invention

The embodiment of the application provides a task management method based on machine learning, which is used for solving the problems that the existing task allocation strategy lacks accuracy, instantaneity and scientificity because the existing task switching management scheme mainly depends on manual setting of a background manager and a task management scheduling mode is single, so that the service efficiency of an intelligent voice interaction robot and an artificial seat is greatly reduced.

The embodiment of the application also provides a task management device based on machine learning, which is used for solving the problem that the existing task dispatching strategy lacks accuracy, instantaneity and scientificity because the existing task switching management scheme mainly depends on the manual setting of a background manager and the task management scheduling mode is single, thereby greatly reducing the service efficiency of the intelligent voice interaction robot and the artificial seat.

The embodiment of the application also provides task management equipment based on machine learning, which is used for solving the problem that the existing task dispatching strategy lacks accuracy, instantaneity and scientificity because the existing task switching management scheme mainly depends on the manual setting of a background manager and the task management scheduling mode is single, thereby greatly reducing the service efficiency of the intelligent voice interaction robot and the artificial seat.

The embodiment of the application also provides a computer readable storage medium, which is used for solving the problem that the existing task allocation strategy lacks accuracy, instantaneity and scientificity because the existing task switching management scheme mainly depends on the manual setting of a background manager and the task management scheduling mode is single, so that the service efficiency of the intelligent voice interaction robot and the artificial seat is greatly reduced.

The embodiment of the application adopts the following technical scheme:

a machine learning based task management method comprising: responding to a received first task, and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identification of the user corresponding to the first task; according to a pre-constructed question library, interacting with the user to obtain question and answer data of the user in the interaction process; processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data; judging whether the first task needs task allocation or not according to the user intention; when the first task is judged to be subjected to task allocation according to the user intention, determining an artificial seat corresponding to the first task, and allocating the first task to the artificial seat for processing.

A machine learning based task management device comprising: the call initiation unit is used for responding to a received first task and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identifier of the user corresponding to the first task; the question and answer data acquisition director is used for interacting with the user according to a pre-constructed question library to acquire question and answer data of the user in the interaction process; the intention recognition unit is used for processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining the intention of a user according to the prediction data; the task management unit is used for judging whether the first task needs task allocation or not according to the user intention; and the task allocation unit is used for determining the artificial seat corresponding to the first task and allocating the first task to the artificial seat for processing when judging that the first task needs to be subjected to task allocation according to the user intention.

A machine learning based task management device comprising:

a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to: responding to a received first task, and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identification of the user corresponding to the first task; according to a pre-constructed question library, interacting with the user to obtain question and answer data of the user in the interaction process; processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data; judging whether the first task needs task allocation or not according to the user intention; when the first task is judged to be subjected to task allocation according to the user intention, determining an artificial seat corresponding to the first task, and allocating the first task to the artificial seat for processing.

A computer-readable storage medium storing one or more programs that, when executed by an electronic device comprising a plurality of application programs, cause the electronic device to: responding to a received first task, and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identification of the user corresponding to the first task; according to a pre-constructed question library, interacting with the user to obtain question and answer data of the user in the interaction process; processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data; judging whether the first task needs task allocation or not according to the user intention; when the first task is judged to be subjected to task allocation according to the user intention, determining an artificial seat corresponding to the first task, and allocating the first task to the artificial seat for processing.

The above-mentioned at least one technical scheme that this application embodiment adopted can reach following beneficial effect:

when the task management method based on machine learning provided by the embodiment of the application is adopted, during task dispatch, tasks can be distributed to the intelligent robot first, the intelligent robot responds to the received first task, after determining the corresponding user according to the user identifier carried in the first task, a voice call request is initiated to the user corresponding to the first task, and during voice interaction, the intelligent robot can conduct question-answer communication with the user according to a pre-constructed question library and acquire question-answer data of the user during interaction; processing the acquired question-answer data according to a pre-trained prediction model to obtain prediction data, determining user intention according to the prediction data, and judging whether the first task needs to be switched to a manual seat according to the user intention; and when the judgment result is yes, the intelligent robot further determines the artificial seat matched with the first task and prepares the first task to the corresponding artificial seat for processing. According to the task management method provided by the scheme, the task (such as sales task) can be executed through the intelligent robot, the intelligent robot can determine the user intention (such as ordering intention) through interaction with the user, when the user is determined to have the corresponding intention (such as purchasing intention), the intelligent robot can switch the first task to the artificial seat for processing, the intelligent robot can accurately screen the intended user, task processing efficiency is greatly improved, and meanwhile, the intelligent robot can switch the first task corresponding to the intended user to the artificial seat for processing. In addition, when the intelligent robot switches the first task to the artificial seat, the artificial seat matched with the user characteristic is selected to perform task allocation according to the user characteristic, so that the success rate of the task is further improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute an undue limitation to the application. In the drawings:

fig. 1 is a specific flow diagram of a task management method based on machine learning according to an embodiment of the present application;

FIG. 2 is a schematic diagram of a specific structure of a decision tree according to an embodiment of the present application;

fig. 3 is a schematic diagram of a specific structure of a red black tree according to an embodiment of the present application;

fig. 4 is a schematic specific structure diagram of a task management device based on machine learning according to an embodiment of the present application;

fig. 5 is a schematic specific structure diagram of a task management device based on machine learning according to an embodiment of the present application.

Detailed Description

For the purposes, technical solutions and advantages of the present application, the technical solutions of the present application will be clearly and completely described below with reference to specific embodiments of the present application and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.

The following describes in detail the technical solutions provided by the embodiments of the present application with reference to the accompanying drawings.

The task management method based on machine learning is used for solving the problem that an existing task dispatching strategy lacks accuracy, instantaneity and scientificity because an existing task switching management scheme mainly depends on manual setting of a background manager and a task management scheduling mode is single, so that the service efficiency of an intelligent voice interaction robot and an artificial seat is greatly reduced.

The execution subject of the machine learning-based task management method provided in the embodiment of the present application may be, but is not limited to, at least one of a task management server, a policy processing server, a sales server, a commodity recommendation server, a shopping server, and the like; the method may be executed by an intelligent robot or an Application (APP) running on the servers.

For convenience of description, embodiments of the method will be described below taking an execution subject of the method as an example of an intelligent robot running on a sales task management server. It will be appreciated that the subject of the method is an exemplary illustration of an intelligent robot and should not be construed as limiting the method.

The specific implementation flow diagram of the task management method based on machine learning provided by the application is shown in fig. 1, and mainly comprises the following steps:

step 11, the intelligent robot responds to a received first task and initiates a voice call request to a user corresponding to the first task;

the first task carries a user identifier of a user corresponding to the first task, and the user identifier can be an identifier which can uniquely identify the user, such as an ID of a user registration account on a system or a user name. For convenience of description, the method provided in the application embodiment will be described below by taking the first task as a sales task as an example. In one embodiment, the sales task may further include task related information such as task processing time and task follow-up node, in addition to the user identifier, so that the intelligent robot may execute a corresponding sales task according to the task related information.

In the embodiment of the application, the first task may be distributed and set by a background management user of the sales system according to service requirements and issued to each intelligent robot on the system; or, the first task may be automatically issued to each intelligent robot by the task scheduling module of the sales system according to a preset rule, and the issuing allocation mode of the first task is not limited in the embodiment of the present application.

It should be noted that, in the embodiment of the present application, the intelligent robot refers to a technical solution such as a semantic retrieval technology based on natural language understanding, a multi-channel knowledge service technology, and a large-scale knowledge base construction technology, and according to a user service requirement, the intelligent robot is trained in advance, and can perform corresponding service (such as voice interaction, service query, service popularization, and fault reporting) in a specified service field (such as a sales service field, a logistics service field, and an after-sales service field). In this embodiment of the present application, the training of the intelligent robot may be performed according to a technical scheme commonly used in the prior art, so the training method related to the intelligent robot is not described herein.

After receiving the first task issued by the sales task management server, the intelligent robot can search a call identifier corresponding to the user identifier in a pre-created user database according to the user identifier carried in the first task, wherein the call identifier can be, for example, a telephone number of a user or an account corresponding to other instant messaging tools of the user, and when the call identifier is the telephone number, the intelligent robot can directly initiate a voice call request to the user by dialing the telephone number; and when the call identifier is an account number corresponding to the timely communication tool, the intelligent robot can directly initiate a voice call request to the account number through a network.

In the embodiment of the application, the intelligent robot specifically may call the user by adopting the following method: determining a call identifier corresponding to the user according to the user identifier carried in the first task; and sending a voice call request to the user according to the call identifier.

Step 12, interacting with the user according to a pre-constructed question library to obtain question and answer data of the user in the interaction process;

the interaction mode of the intelligent robot and the user is a main factor for determining whether the sales task can be continued, in order to improve the user experience degree in the interaction process and improve the order forming completion rate of the sales task, the intelligent robot can inquire the user characteristics according to the user identification in the voice interaction process with the user, further find the most matched interaction problem with the user in a pre-built problem library according to the user characteristics, and perform voice interaction with the user based on the interaction problems. In the embodiment of the present application, the specific implementation manner of step 12 may include: determining the user characteristics according to the user identification; according to the user characteristics, determining interaction problems corresponding to the user in the pre-constructed problem library; and according to the interaction problem, interacting with the user.

In one embodiment, the sales task management server may obtain a sales record corresponding to a sales task completed in a history, where the sales record stores a dialogue record between a human agent (or an intelligent robot) and a user identifier of the user, so that the sales task management server may determine a key attention question of the user by performing semantic analysis on the dialogue record in the sales record, determine a user feature according to the user identifier, further determine attention questions corresponding to different user features by analyzing a massive amount of history, and further generate a "user feature-attention question correspondence table" according to the focus question, and then in a subsequent use process, the intelligent robot may determine attention questions corresponding to different users through the "user feature-attention question correspondence table", further determine an interaction question corresponding to the user in a pre-constructed question library according to the attention question, perform voice interaction with the user based on the interaction questions, and collect answer data of the user for the interaction question in the interaction process for analysis and use in a subsequent step.

Specifically, the sales task management server may create a user database for users who have used the system (such as purchased products or opened related services on the sales system), where related information of the users, such as age, work, city, gender, and historical consumption conditions of the users, etc., and extract user features according to the related information of the users, so as to generate a user feature database. Subsequently, when the intelligent robot initiates a voice call to the user in response to the sales task, the intelligent robot may first search for the user feature of the user in a pre-built user feature database according to the user identifier, and then match the corresponding interaction problem in the problem library according to the searched user feature.

In addition, it should be noted that, when the user targeted by the sales task is a new user (i.e. a user who has not used the sales system), since the sales system does not store the feature data corresponding to the user, the intelligent robot can search the problem library for the general interaction problem, and complete the voice interaction with the user using the general interaction problem. In the embodiment of the application, the universal interaction problem can be set by a background manager of the sales system according to the historical service condition.

Step 13, processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data;

in one embodiment, the prediction model may refer to a decision tree prediction model constructed by training with a pre-collected training sample based on a decision tree algorithm, and in this embodiment of the present application, the decision tree prediction model may be specifically obtained by training in the following manner, which specifically includes the following sub-steps:

sub-step 1301, determining information entropy data corresponding to a training sample by utilizing a decision tree algorithm according to a question library and the training sample;

assume that there are n questions in the pre-constructed question bank, denoted as { q } ₁ ,q ₂ ...q _n In the embodiment of the present application, the question library may be the question library used in step 12, or may be a question library formed by a part of questions screened from the question library used in step 12 according to service requirements. Assume that there are a total of x sample users in the pre-constructed training sample library D, denoted as { D ] ₁ ,d ₂ ...d _x Training sample data may be represented by table 1 below, where "Y" or "N" represents answer data made by each sample for each question in the question bank; "intent" indicates whether the sample users have an intent to order the sales task.

	q ₁	q ₂	……	q _n	Intent (R)
						d ₁	Y	N	……	N	Has the following components
d ₂	N	Y	……	Y	Has the following components
						……	……	……	……	……	……
d _x	N	Y	……	Y	Without any means for

TABLE 1

In the embodiment of the present application, according to the decision tree algorithm, the "information entropy" of each sample may be determined by calculation according to the following formula [1 ]:

wherein P (d) _i ) Representing the corresponding meaning in the sampleThe probability of occurrence in all intent results of the whole sample set is given to (R), for example, assuming that there are 10 samples in total in the above sample set. The intent result of the ith sample is "have", there are 6 "none" and 4 "have" in the sample set, then P (d _i ) 40%, i.e. 0.4.

Sub-step 1302, determining condition entropy data corresponding to a training sample by utilizing a decision tree algorithm according to a question library and the training sample;

In the embodiment of the present application, the value of each problem in the sample data table may be taken as a condition, and then "condition entropy" data of the sample relative to the condition may be calculated according to the following formula [2 ]:

wherein, the liquid crystal display device comprises a liquid crystal display device,

for questions in sample data>

The probability of each value appearing (together +.>

Seed value),

to solve the problem q _j Under a certain value condition, the result d appears _i Is a probability of (2). For example, in 10 samples, problem q _j The number of "Y" is 3, then +.>

30%, i.e. 0.3. Suppose d _i The corresponding value is "have", at q _j Of the 3 results of "Y", the number of "intent (R)" being "have" is 2, then +.>

2/3, i.e. 0.67.

Sub-step 1303, constructing a first prediction model according to the information entropy data and the conditional entropy data obtained by executing the calculation in the above step;

in the embodiment of the application, the information gain corresponding to each sample can be calculated through the following formula [3], and then a decision tree prediction model is constructed according to the information gain:

Gain(D|Q＝q _j )＝H(D)-H(D|Q＝q _j )[3]

the "information gain" for each question in the question bank can be calculated according to the calculation of equation [3 ].

A sub-step 1304 of performing predictive training on the training sample according to the first predictive model obtained by executing the sub-step 1303 to obtain the predictive model;

In the embodiment of the application, the problem of maximum information gain can be selected as a root node, then a sample is tested according to the problem, and the sample is divided into different subsets according to different values; if the subset has only one corresponding result value, the subset branches to serve as leaf nodes; otherwise, recursion of the algorithm above on the sub-sets results in a decision tree as shown in FIG. 2.

In one embodiment, in order to improve the accuracy of the prediction model, s sample sets may be randomly extracted in a random sampling manner, and s decision trees are respectively calculated for the s sample sets to form a decision forest with s decision trees in total, and then the decision tree forest is used as the prediction model.

Step 14, judging whether to perform task allocation for the first task according to the user intention obtained by executing the step 13, executing the step 15 when the judgment result is that task allocation is required, and executing the step 16 when the judgment result is that task allocation is not required;

in this embodiment of the present application, because the artificial agent can more accurately complete communication with the user compared with the intelligent robot, the possibility of completing the sales task can be greatly improved, but compared with the intelligent robot, the processing efficiency of the artificial agent is lower, so in one embodiment, the sales task corresponding to the intended user can be switched to the artificial agent, so as to improve the success rate of the sales task, and in this embodiment of the present application, the intelligent robot can specifically determine whether to perform task switching according to the following method: when the user intention is determined to be yes, task allocation is determined to be carried out for the first task; and when the user intention is determined to be NO, determining that task allocation is not performed for the first task.

Step 15, when the judgment result obtained by executing the step 14 is yes, the intelligent robot can further determine an artificial seat corresponding to the first task, and allocate the first task to the artificial seat for processing;

here, before the intelligent robot switches the sales task to the manual agent processing, the intelligent robot may match the optimal task processor in the manual agent according to the user characteristics and the question-answer data collected by executing the step 12, and specifically, in one embodiment, the intelligent robot may determine the optimal processing corresponding to the sales task according to the following method: "Artificial seat": acquiring user characteristics of the user according to the user identification carried in the first task; and acquiring the seat characteristics of the artificial seat, and determining the artificial seat corresponding to the first task according to the matching degree of the seat characteristics and the user characteristics.

In one embodiment, the user characteristics may include: gender, age, user quality score, whether married, child family, etc., assuming n features, the customer is denoted as c, the above features are respectively denoted as c _i In addition, the collected related data of the communication process between the intelligent robot and the user can be used as the characteristics of the user, for example, the communication duration t of the user can be used as one of the characteristics of the client, and then the characteristic set of the client c can be expressed as follows: { c ₁ ,c ₂ ....c _n ,t}。

In addition, in the embodiment of the present application, the manual agent feature library mainly includes: sex, age group, service (corresponding to user's quality score), professional (corresponding to user's occupation), marital status, child's family success rate, etc., the features of the artificial agent and the user's feature attributes are in one-to-one correspondence, and are obtained based on a certain set of user samples C for statistical training. In the embodiment of the application, the process of statistical training for the artificial seat features is as follows:

taking the success rate of the continuous protection as an example, assume that the event of the success rate of the continuous protection of the user sample is set to be s, and all attribute conditions { c) of the user c are based according to a naive Bayesian algorithm ₁ ,c ₂ ....c _n T are independent of each other, and the attribute c for the user c is obtained according to training sample statistics _i Success rate of P (c) _i |s)

For discrete attributes, e.g. let attribute c _i Representing whether the married person is married or not, assuming that the training sample of the artificial agent is 100 clients, wherein the number of the married clients is 31, the artificial agent corresponds to the attribute c _i Success rate P (c) _i S) is 31%, i.e., 0.31, and the success rate is taken as a characteristic value of the characteristic of the manual agent 'user marital status good length'.

In the embodiment of the application, the feature library of the manual agent mainly stores all the attributes c of the user obtained based on sample training _i Success rate set { P (c) ₁ |s)P(c ₂ |s)......P(c _n S), P (t s). For client c, the total power may be calculated according to naive Bayesian principle by the following equation [4]And (3) determining:

/>

according to the formula [4], the success rate of each artificial agent facing the corresponding sales tasks of the user can be calculated, and the first x agents are selected as candidate artificial agents according to the success rate ranking.

After considering the success rate of the task, before further selecting the artificial agents to perform task switching, in order to give the artificial agents a relatively average workload and a relatively fair opportunity, the evaluation functions of the candidate artificial agents can be considered, and the candidate artificial agents are prioritized according to the evaluation functions, so that the most suitable artificial agents are screened out to perform task switching. In one embodiment, the evaluation value of each artificial agent may be determined according to the following equation [5 ]:

for team manager to give artificial seat m _i And (5) setting a regulating factor. />

Is a manual seat m _i Time length from last call answering, +.>

Is a manual seat m _i Cumulative total time of answering call on the same day, +. >

Is a manual seat m _i Total number of calls received on the same day,/->

Is a manual seat m _i Sales are made on the same day.

For all the artificial agents participating in the voice response in the same day, the evaluation value calculated according to the evaluation function can be used as a sorting value to generate a red black tree for storing the sorting value of all the artificial agents. For example, taking 6 nodes as an example, a red-black tree is shown in fig. 3, the red-black tree is a binary search tree, and coloring is added on the basis of the binary search tree. The method is characterized by mainly comprising the following steps:

a. if the left subtree of a certain node is not empty, the values of all nodes on the left subtree are smaller than the value of the root node of the node;

b. if the right subtree of a certain node is not empty, the values of all nodes on the right subtree are larger than the value of the root node of the node;

c. each node is either black or red, and the root node is black, and each leaf node (NIL node) is black.

d. If a node is red, its two child sub-nodes must be black.

e. All simple paths from any one node to its leaf node contain the same number of black nodes.

f. The left and right subtrees of any node also recursively conform to the above features.

A height of at most 2log (n+1) of a black tree containing n nodes, and a time complexity of search, insertion, deletion of the black tree of O (log n), because the medium-order traversal of the black tree is ordered, according to an evaluation function F (m _i ) The values of (2) are ordered from small to large so that the first artificial agent appearing, i.e. the best matching artificial agent to answer, is found only along the mid-order traversal direction of the mangrove.

After the best matched artificial seat is selected by the method, the intelligent robot can press the sales task to the artificial seat, simultaneously switch the voice call with the user to the artificial seat in real time, and answer and continue to serve the user by the artificial seat.

In one embodiment, the intelligent robot may also forward the dialogue record with the user to the artificial agent, so that the artificial agent can know the previous processing progress according to the dialogue record.

And step 16, when the judgment result obtained by executing the step 14 is negative, the intelligent robot can continue to carry out voice call with the user until the call is ended.

In an implementation manner, the embodiment of the application also provides a task management device based on machine learning, which is used for solving the problem that the existing task dispatching strategy lacks accuracy, instantaneity and scientificity because the existing task switching management scheme mainly depends on the manual setting of a background manager and the task management scheduling mode is single, so that the service efficiency of the intelligent voice interaction robot and the manual seat is greatly reduced. The specific structure diagram of the access flow control device is shown in fig. 4, and includes: a call initiation unit 41, a question and answer data acquisition unit 42, an intention recognition unit 43, a task management unit 44, and a task orchestration unit 45.

The call initiation unit 41 is configured to initiate a voice call request to a user corresponding to a first task in response to the received first task, where the first task carries a user identifier of the user corresponding to the first task;

a question and answer data acquisition unit 42, configured to interact with the user according to a pre-constructed question library, and acquire question and answer data of the user in the interaction process;

an intention recognition unit 43, configured to process the question-answer data according to a pre-trained prediction model, obtain prediction data, and determine a user intention according to the prediction data;

A task management unit 44, configured to determine whether to perform task allocation for the first task according to the user intention;

and the task allocation unit 45 determines an artificial seat corresponding to the first task and allocates the first task to the artificial seat for processing when the judgment result is yes.

In one embodiment, the current limiting unit 23 is specifically configured to: determining a first access token pool corresponding to the target access address, and judging whether the number of access tokens in the first access token pool is larger than the first number; when the number of the access tokens in the first access token pool is judged to be larger than the first number, determining a second access token pool corresponding to the service type to be accessed; determining whether the number of access tokens in the second pool of access tokens is greater than the first number.

In one embodiment, the call initiation unit 41 is configured to determine, according to a user identifier carried in the first task, a call identifier corresponding to the user; and sending a voice call request to the user according to the call identifier.

In one embodiment, the question and answer data collecting unit 42 is configured to determine the user characteristic according to the user identifier; according to the user characteristics, determining interaction problems corresponding to the user in the pre-constructed problem library; and according to the interaction problem, interacting with the user.

In one embodiment, the system further comprises a prediction model training unit, which is used for determining information entropy data and conditional entropy data corresponding to a training sample by utilizing a decision tree algorithm according to the problem library and the training sample; constructing a first prediction model according to the information entropy data and the conditional entropy data; and carrying out predictive training on the training sample according to the first predictive model to obtain the predictive model.

In one embodiment, the task management unit 44 is configured to, when determining that the user intention is yes, determine to perform task allocation for the first task; and when the user intention is determined to be NO, determining that task allocation is not performed for the first task.

In one embodiment, the task allocation unit 45 is configured to obtain a user characteristic of the user according to a user identifier carried in the first task; and acquiring the seat characteristics of the artificial seat, and determining the artificial seat corresponding to the first task according to the matching degree of the seat characteristics and the user characteristics.

Fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application. Referring to fig. 5, at the hardware level, the electronic device includes a processor, and optionally an internal bus, a network interface, and a memory. The Memory may include a Memory, such as a Random-Access Memory (RAM), and may further include a non-volatile Memory (non-volatile Memory), such as at least 1 disk Memory. Of course, the electronic device may also include hardware required for other services.

The processor, network interface, and memory may be interconnected by an internal bus, which may be an ISA (Industry Standard Architecture ) bus, a PCI (Peripheral Component Interconnect, peripheral component interconnect standard) bus, or EISA (Extended Industry Standard Architecture ) bus, among others. The buses may be classified as address buses, data buses, control buses, etc. For ease of illustration, only one bi-directional arrow is shown in FIG. 5, but not only one bus or type of bus.

And the memory is used for storing programs. In particular, the program may include program code including computer-operating instructions. The memory may include memory and non-volatile storage and provide instructions and data to the processor.

The processor reads the corresponding computer program from the nonvolatile memory into the memory and then runs, and forms a task management device for machine learning on a logic level. The processor is used for executing the programs stored in the memory and is specifically used for executing the following operations:

responding to a received first task, and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identification of the user corresponding to the first task; according to a pre-constructed question library, interacting with the user to obtain question and answer data of the user in the interaction process; processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data; judging whether task allocation is carried out for the first task according to the user intention; and when the judgment result is yes, determining the artificial seat corresponding to the first task, and allocating the first task to the artificial seat for processing.

The method performed by the machine learning based task management electronic device as disclosed in the embodiment shown in fig. 5 of the present application may be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or by instructions in the form of software. The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but also digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), field programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present application may be embodied directly in hardware, in a decoded processor, or in a combination of hardware and software modules in a decoded processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory, and the processor reads the information in the memory and, in combination with its hardware, performs the steps of the above method.

Of course, other implementations, such as a logic device or a combination of hardware and software, are not excluded from the electronic device of the present application, that is, the execution subject of the following processing flow is not limited to each logic unit, but may be hardware or a logic device.

The present embodiments also provide a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a portable electronic device comprising a plurality of application programs, enable the portable electronic device to perform the method of the embodiment of fig. 1, and in particular to:

It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.

Claims

1. A machine learning based task management method, comprising:

responding to a received first task, and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identification of the user corresponding to the first task;

According to a pre-constructed question library, interacting with the user to obtain question and answer data of the user in the interaction process;

processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining user intent according to the prediction data;

judging whether task allocation is carried out for the first task according to the user intention;

and when the judgment result is yes, determining the artificial seat corresponding to the first task, and allocating the first task to the artificial seat for processing.

2. The method according to claim 1, wherein the responding to the received first task initiates a voice call request to the user corresponding to the first task, specifically comprises:

determining a call identifier corresponding to the user according to the user identifier carried in the first task;

and sending a voice call request to the user according to the call identifier.

3. The method according to claim 1, wherein the interaction with the user is performed according to a pre-constructed question bank, in particular comprising:

determining the user characteristics according to the user identification;

According to the user characteristics, determining interaction problems corresponding to the user in the pre-constructed problem library;

and according to the interaction problem, interacting with the user.

4. The method according to claim 1, characterized in that said pre-trained predictive model comprises in particular:

determining information entropy data and condition entropy data corresponding to the training samples by utilizing a decision tree algorithm according to the problem library and the training samples;

constructing a first prediction model according to the information entropy data and the conditional entropy data;

and carrying out predictive training on the training sample according to the first predictive model to obtain the predictive model.

5. The method according to claim 1, wherein the determining whether to perform task allocation for the first task according to the user intention specifically includes:

when the user intention is determined to be yes, task allocation is determined to be carried out for the first task;

and when the user intention is determined to be NO, determining that task allocation is not performed for the first task.

6. The method according to claim 1, wherein the determining the artificial agent corresponding to the first task specifically comprises:

Acquiring user characteristics of the user according to the user identification carried in the first task;

and acquiring the seat characteristics of the artificial seat, and determining the artificial seat corresponding to the first task according to the matching degree of the seat characteristics and the user characteristics.

7. A machine learning based task management device, comprising:

the call initiation unit is used for responding to a received first task and initiating a voice call request to a user corresponding to the first task, wherein the first task carries a user identifier of the user corresponding to the first task;

the question and answer data acquisition unit is used for interacting with the user according to a pre-constructed question library to acquire question and answer data of the user in the interaction process;

the intention recognition unit is used for processing the question-answer data according to a pre-trained prediction model to obtain prediction data, and determining the intention of a user according to the prediction data;

the task management unit is used for judging whether task allocation is carried out for the first task according to the user intention;

and the task allocation unit is used for determining the artificial seat corresponding to the first task and allocating the first task to the artificial seat for processing when the judgment result is yes.

8. The apparatus according to claim 7, further comprising a predictive model training unit, in particular for:

9. A machine learning based task management device comprising:

a processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the method of any one of claims 1 to 6.

10. A computer readable storage medium, characterized in that instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the method of any one of claims 1 to 6.