CN113570324A - Outbound call flow editing method, device, electronic device and storage medium - Google Patents
Outbound call flow editing method, device, electronic device and storage medium Download PDFInfo
- Publication number
- CN113570324A CN113570324A CN202110636513.1A CN202110636513A CN113570324A CN 113570324 A CN113570324 A CN 113570324A CN 202110636513 A CN202110636513 A CN 202110636513A CN 113570324 A CN113570324 A CN 113570324A
- Authority
- CN
- China
- Prior art keywords
- node
- sample
- type
- attribute
- outbound call
- 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 OR CALCULATING; 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
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06Q30/0281—Customer communication at a business location, e.g. providing product or service information, consulting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Economics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Finance (AREA)
- Biophysics (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本发明提供一种外呼流程编辑方法、装置、电子设备及存储介质，其中方法包括：获取待编辑外呼流程中当前节点的节点类型和节点属性；将所述当前节点的节点类型和节点属性输入至节点预测模型，得到所述节点预测模型输出的下一节点的候选节点；基于所述候选节点的节点类型和节点属性，确定所述待编辑外呼流程中的下一节点；其中，所述节点属性包括节点提示音话术，所述节点预测模型是基于样本外呼流程和/或样本节点提示音话术训练得到的。本发明提供的方法、装置、电子设备及存储介质，降低了外呼流程编辑的难度，提高了外呼流程编辑的效率，所得到的外呼流程中各个节点的关联性强，实用性强。
The present invention provides a method, device, electronic device and storage medium for editing an outbound call process, wherein the method includes: acquiring the node type and node attribute of the current node in the outbound call process to be edited; Input to the node prediction model, and obtain the candidate node of the next node output by the node prediction model; based on the node type and node attribute of the candidate node, determine the next node in the outbound call process to be edited; wherein, the The node attribute includes the node prompting voice, and the node prediction model is obtained by training based on the sample outbound call flow and/or the sample node prompting voice. The method, device, electronic device and storage medium provided by the invention reduce the difficulty of editing the outbound call process, improve the efficiency of the outbound call process editing, and the obtained outbound call process has strong correlation and practicability.
Description
Technical Field
The invention relates to the technical field of computers, in particular to an outbound flow editing method and device, electronic equipment and a storage medium.
Background
The outbound refers to that the customer service center system actively initiates a call to the customer to provide information consultation service for the customer. The outbound flow can be established in advance through a simple and visual graphical editing mode, and the efficiency of outbound conversation is effectively improved. The outbound flow comprises a plurality of flow nodes which need to communicate with the client in a service scene and attributes corresponding to the flow nodes.
In the prior art, node selection and attribute configuration are automatically completed by service personnel in the outbound flow, so that the service personnel are required to have rich service experience and master professional flow editing knowledge. The existing outbound flow editing method has the disadvantages of high editing difficulty of the outbound flow, low editing efficiency and poor practicability of the obtained outbound flow.
Disclosure of Invention
The invention provides an outbound flow editing method, an outbound flow editing device, electronic equipment and a storage medium, which are used for solving the technical problems that the outbound flow in the prior art is high in editing difficulty, low in editing efficiency and poor in practicability of the obtained outbound flow.
The invention provides an outbound flow editing method, which comprises the following steps:
acquiring the node type and node attribute of a current node in an outbound flow to be edited;
inputting the node type and the node attribute of the current node into a node prediction model to obtain a candidate node of a next node output by the node prediction model;
determining a next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node;
the node attributes comprise node prompt phonetics, and the node prediction model is obtained based on a sample outbound flow and/or sample node prompt phonetics training.
According to the outbound flow editing method provided by the invention, the node prediction model is obtained by training based on the following steps:
determining a plurality of sample outbound flows and a plurality of sample node prompt phonetics;
determining the association degree between nodes of different node types in each sample outbound flow and the evaluation label of each sample node prompt phonetics; the evaluation label is used for representing the semantic expression capability of the sample node prompt phonetics;
and training an initial model based on the correlation degrees among the nodes of different node types in the sample outbound flows and each sample outbound flow and the evaluation labels of the sample node prompt phonetics and each sample node prompt phonetics to obtain the node prediction model.
According to the outbound flow editing method provided by the invention, the step of determining the next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node comprises the following steps:
determining client intention prediction information based on the node type and/or node attribute of the current node;
and determining the similarity between the client intention prediction information and the node prompt phonetics in the node attributes of the candidate nodes, and taking the candidate node with the highest similarity as the next node in the outbound flow to be edited.
According to the outbound flow editing method provided by the invention, the step of determining the next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node comprises the following steps:
determining the prediction weight of each node type based on the number of the nodes of each node type in the sample outbound flow and the total number of the nodes of the sample outbound flow;
and determining the next node in the outbound flow to be edited based on the prediction weight of each node type and the node type of the candidate node.
According to the outbound flow editing method provided by the invention, the step of determining the next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node comprises the following steps:
determining the correlation between the node attribute of each candidate node and the node attribute of the current node based on the node type and the node attribute of each candidate node, the node type and the node attribute of the current node and a preset knowledge graph;
determining a next node in the outbound flow to be edited based on the correlation between the node attribute of each candidate node and the node attribute of the current node;
wherein the preset knowledge-graph is determined based on the nodes of different node types and the correlation between the node attributes of the nodes of different node types.
According to the outbound flow editing method provided by the invention, the step of determining the next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node comprises the following steps:
determining a precision enhancement sample based on the node type and the node attribute of the next node and the node type and the node attribute of the current node;
and performing optimization training on the node prediction model based on the precision enhancement sample.
According to the outbound flow editing method provided by the invention, the node type comprises at least one of a request input node, an expected node, a judgment node and a prompt tone node; the node attribute further comprises a node name and a dialect type;
the request input node is used for sending input prompt information to the client and determining a corresponding slot position based on the input prompt information, wherein the slot position is used for storing the input information of the client;
the expected node is used for acquiring input information of a client based on the slot position;
the judgment node is used for determining the selection result of the customer based on the input information of the previous node and a preset given value;
the prompt tone node is used for sending prompt information to the client based on the set prompt type, and the prompt type comprises text and/or voice.
The invention also provides an outbound flow editing device, comprising:
the obtaining unit is used for obtaining the node type and the node attribute of the current node in the outbound flow to be edited;
the prediction unit is used for inputting the node type and the node attribute of the current node into a node prediction model to obtain a candidate node of a next node output by the node prediction model;
a determining unit, configured to determine a next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node;
the node attributes comprise node prompt phonetics, and the node prediction model is obtained based on a sample outbound flow and/or sample node prompt phonetics training.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program which is stored on the memory and can be run on the processor, wherein the processor realizes the steps of the outbound flow editing method when executing the program.
The invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the outbound flow editing method.
The method, the device, the electronic equipment and the storage medium for editing the outbound flow input the node type and the node attribute of the current node in the outbound flow to be edited to the node prediction model to obtain a plurality of candidate nodes of the next node, so as to determine the next node corresponding to the current node, wherein the node prediction model is obtained based on sample outbound flow and sample node prompt phonetics training, can accurately predict the node in the outbound flow to be edited and the prompt phonetics of the node, guides a service worker to edit the outbound flow, reduces the difficulty of the outbound flow editing, improves the efficiency of the outbound flow editing, and has strong relevance and practicability of each node in the obtained outbound flow.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are 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 flow chart of an outbound flow editing method according to the present invention;
FIG. 2 is a second schematic flow chart of the outbound flow editing method according to the present invention;
FIG. 3 is a schematic structural diagram of an outbound flow editing apparatus according to the present invention;
fig. 4 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the prior art, tools for calling out a flow and a using method thereof need business personnel to have abundant business experience and master professional flow editing knowledge. The business personnel experience is poor, for example, the business personnel are not familiar with how to edit the flow, so that the business personnel often have no way to determine the outbound flow, do not know what the next node should receive, and do not know how the variable scope is defined. Even if the outbound flow is edited, the outbound flow can not be used normally. In addition, when the prompt tone nodes are edited, the problem of how to write the dialogs is often faced, and a great deal of time and energy are wasted.
In view of the deficiency of the prior art, fig. 1 is a schematic flow diagram of an outbound flow editing method provided by the present invention, as shown in fig. 1, the method includes:
Specifically, the outbound flow is a conversation flow set according to a specific business scenario and providing information consultation service for the client.
The outbound process requires the service personnel to edit according to the transaction sequence of the items in the service scene, including a plurality of nodes to be communicated with the client and the node attributes corresponding to each node. The nodes in the outbound flow are used for sending information requests to clients, playing prompt tones to the clients or carrying out logic judgment on information input by the clients and the like.
The node type is a functionally classified category of the node. For example, the node type is a node requesting input, and functions to issue an information input request to a client, from which a user is expected to input corresponding information. A node attribute is a property or characteristic of a node. For example, for a node whose node type is a request input, its node attributes may include a node name, a node alert phonetics, a phonetics type, and so on.
Since the outbound flow needs to communicate with the client, the node attributes of each node type include node-prompted phonetics. The node prompt phonetics are dialog content used to provide prompt information to the client.
For example, one outbound flow in the call charge query may be:
and the node 1:
the node type is as follows: request input node
Node prompt phonetics: asking if you want to inquire the charge "
Nodal-conversational type: speech sound
And (3) the node 2:
the node type is as follows: request input node
Node prompt phonetics: "Please input the month you want to inquire"
Nodal-conversational type: speech sound
And (3) the node:
the node type is as follows: information output node
Node prompt phonetics: "the call charge of the month you inquire is XX Yuan"
Nodal-conversational type: speech sound
The outbound flow to be edited may only include the node 1 therein, and the node 2 needs to be predicted according to the node 1.
And 120, inputting the node type and the node attribute of the current node into the node prediction model to obtain a candidate node of the next node output by the node prediction model. The node prediction model is obtained based on sample outbound flow and/or sample node prompt phonetics training.
Specifically, for the current node, the next node is not fixed and may correspond to multiple candidate nodes. For example, the current node is a request input node "please input the type of service you want to query", and if the customer feedback is "telephone charge", the next node may be a request input node "please input the month you want to query"; if the feedback opinion of the client is 'do not want to inquire', the feedback opinion can also be 'thank you for incoming call' of the prompt tone node.
And predicting the next node of the current node according to the node type and the node attribute of the current node through the node prediction model to obtain a plurality of candidate nodes. Each candidate node maintains coherence with the current node in the service processing flow, that is, each candidate node can be used as a node next to the current node.
For example, the node prediction model may adopt a neural network model, and specifically includes a feature extraction layer, a feature prediction layer, and a result output layer. The feature extraction layer is used for extracting features of node types and node attributes of the current node, the feature prediction layer is used for learning according to the extracted features of the current node, analyzing and judging a plurality of candidate nodes which are closely related to the current node in a business processing flow, and the result output layer is used for determining the possibility that each candidate node becomes a next node according to the association close degree of each candidate node and the current node and outputting each candidate node from large to small.
The node prediction model is mainly used for predicting the node type of the next node and the node prompt phonetics in the node attribute. Thus, it can be trained using sample outbound procedures and sample node prompt phonetics.
The sample outbound flow is multi-round dialogue data containing outbound prompt information and customer response information, and can be used for training the node prediction model and improving the prediction capability of the node prediction model on the node type of the next node.
The initial model can be trained independently by using a sample outbound process to obtain a node prediction model, and the node prediction model can be obtained by the following training mode: first, a large number of sample outbound flows are collected. Secondly, marking each node in each sample outbound flow, and determining the association degree between nodes of different node types. And then, training the initial model according to a large number of sample outbound processes, so that the initial model can learn and calculate the association degree between nodes of different node types, thereby improving the prediction capability of the initial model on the node type of the next node and obtaining a node prediction model.
The sample node prompting phonetics are various statements of used node prompting phonetics or information communication with clients, and can be used for training the node prediction model and improving the expression capability of the node prediction model on the node prompting phonetics of the next node.
The initial model can be trained separately by using a sample node prompt phonetics to obtain a node prediction model, and the method can be specifically implemented by the following training modes: first, a large number of sample node prompt phonetics are collected. And secondly, determining an evaluation label of the prompt phonetics of each sample node according to the semantic expression degree of the prompt phonetics of each sample node. For example, the sample node hinting phonetics with higher semantic expression clarity are set as positive evaluation labels, and the sample node hinting phonetics with lower semantic expression clarity are set as negative evaluation labels. And then, training the initial model according to a large number of sample node prompt phonetics, improving the expression capability of the initial model to the node prompt phonetics of the nodes, and obtaining a node prediction model.
The sample outbound flow and the sample node prompt phonetics can be input into the initial model for training at the same time or can be input into the initial model for training respectively.
Specifically, according to the node type and node attribute of each candidate node, a service person may select and determine the next node according to the actual situation, or may determine the next node according to the prediction score of each candidate node output by the node prediction model. The prediction score of each candidate node is used for measuring the possibility of the candidate node becoming the next node, and can be determined according to the association degree of each candidate node and the current node.
According to the outbound flow editing method provided by the embodiment of the invention, the node type and the node attribute of the current node in the outbound flow to be edited are input into the node prediction model to obtain a plurality of candidate nodes of the next node, so that the next node corresponding to the current node is determined, the node prediction model is obtained based on the sample outbound flow and sample node prompt phonetics training, the node in the outbound flow to be edited and the prompt phonetics of the node can be accurately predicted, a service worker is guided to carry out outbound flow editing, the difficulty of outbound flow editing is reduced, the outbound flow editing efficiency is improved, and the obtained outbound flow has strong relevance and practicability of each node.
Based on the above embodiment, the node prediction model is obtained based on the following training steps:
determining a plurality of sample outbound flows and a plurality of sample node prompt phonetics;
determining the association degree between nodes of different node types in each sample outbound flow and the evaluation label of each sample node prompt phonetics; the evaluation label is used for representing the semantic expression capability of the sample node prompt phonetics;
and training the initial model based on the correlation degree between the nodes of different node types in the sample outbound flows and each sample outbound flow and the evaluation labels of the prompt phonetics of the sample nodes and each sample node to obtain a node prediction model.
Specifically, the embodiment of the present invention provides a joint training method, which may use a sample outbound process and a sample node prompt phonetics to perform joint training on an initial model to obtain a node prediction model, and may specifically use the following training modes:
first, a large number of sample outbound flows and a large number of sample node prompt phonetics are collected. Secondly, after the node types of all the nodes in each sample outbound flow are analyzed, the association degree between the nodes of different node types in each sample outbound flow is determined, and the evaluation label of the prompt phonetics of each sample node is determined according to the semantic expression degree of the prompt phonetics of each sample node. Then, training the initial model according to the correlation degree between the nodes of different node types in the sample outbound flows and each sample outbound flow, the prompt phonetics of the sample nodes and the evaluation label of the prompt phonetics of each sample node, and improving the prediction capability of the initial model on the node type of the next node and the semantic expression capability of the initial model on the prompt phonetics of the node of the next node to obtain a node prediction model.
The degree of association between nodes of different node types may be a degree of similarity between node attributes, for example, if the degree of similarity between node attributes of nodes of two different node types is higher, the degree of association between nodes of the two different node types is higher.
The evaluation label of the sample node prompt phonetics can be determined according to the semantic expression capability of the sample node prompt phonetics. For example, the evaluation label of the sample node prompt phonetics with stronger semantic expression capability may be positive, and the evaluation label of the sample node prompt phonetics with weaker semantic expression capability may be negative.
According to the outbound flow editing method provided by the embodiment of the invention, the node prediction model is obtained through the sample outbound flow and the sample node prompt phonetics combined training, so that the prediction accuracy of the node prediction model is improved.
Based on any of the above embodiments, step 130 includes:
determining client intention prediction information based on the node type and/or node attribute of the current node;
and determining the similarity between the client intention prediction information and the node prompt phonetics in the node attributes of the candidate nodes, and taking the candidate node with the highest similarity as the next node in the outbound flow to be edited.
In particular, the client will be the willingness of the client to respond to the information of the current node. The client intention prediction information is prediction information about client intention made by service personnel. The service personnel can make assumptions about the possible willingness of the client to respond according to the node attributes of the current node, and determine client willingness prediction information according to the assumed results. For example, if the current node is a request input node "please input the month that you want to query", the business person can make a prediction that the customer is about to input the month information, and at this time, the customer will predict that the information is the month that the customer inputs, the node prompts that the desired node in the phonetics including the input month information can be used as the next node.
The next node may be determined from the candidate nodes by a similarity matching method. For example, the client intention prediction information and the node prompting phonetics of each candidate node may be converted into texts in advance, and the similarity between the client intention prediction information and the node prompting phonetics in the node attribute of each candidate node may be determined by a text similarity matching method. And taking the candidate node with the highest similarity as the next node in the outbound flow to be edited.
The text similarity matching may select at least one of a keyword matching, TF-IDF (word frequency-inverse text frequency), euclidean distance, cosine similarity, and the like.
According to the outbound flow editing method provided by the embodiment of the invention, the next node is determined by determining the similarity between the client intention prediction information and the node prompt phonetics in the node attribute of each candidate node, so that the outbound flow editing efficiency is improved, the obtained outbound flow has strong relevance of each node, and the practicability is strong.
Based on any of the above embodiments, step 130 includes:
determining the prediction weight of each node type based on the number of nodes of each node type in the sample outbound flow and the total number of nodes in the sample outbound flow;
and determining the next node in the outbound flow to be edited based on the prediction weight of each node type and the node type of the candidate node.
Specifically, the number of nodes of each node type in the multiple sample outbound flows may be counted, and the percentage of the number of nodes of each node type in the total number of nodes in the sample outbound flows is determined. The higher the percentage is, the higher the probability that the node of the node type appears is, and the higher the possibility of becoming the next node of the current node is; the lower the percentage, the less probability that a node indicating the node type appears, the lower the likelihood of becoming the next node to the current node.
The prediction weight of each node type may be determined according to the percentage of each node type. The higher the percentage, the higher the prediction weight; the lower the percentage, the lower the prediction weight. For example, if the percentage of the requesting input node is higher than the percentage of the desired node, the prediction weight of the requesting input node in the candidate nodes is greater than the prediction weight of the desired node. The next node may be the request input node.
According to the outbound flow editing method provided by the embodiment of the invention, the prediction weight of each node type is determined according to the number of the nodes of each node type in the sample outbound flow, the next node is determined, the outbound flow editing efficiency is improved, the obtained outbound flow has strong relevance of each node, and the practicability is strong.
Based on any of the above embodiments, step 130 includes:
determining the correlation between the node attribute of each candidate node and the node attribute of the current node based on the node type and the node attribute of each candidate node, the node type and the node attribute of the current node and a preset knowledge graph;
determining a next node in the outbound flow to be edited based on the correlation between the node attribute of each candidate node and the node attribute of the current node;
wherein the preset knowledge graph is determined based on the nodes of different node types and the correlation between the node attributes of the nodes of different node types.
Specifically, the correlation between node attributes refers to the degree of association between node attributes. For example, when the node attributes are node cue phonetics, then the correlation between the node attributes may be the degree to which the node cue phonetics are semantically logically related. The node requesting the input node a prompts the phonetics as "please input the month you want to inquire", and the node requesting the node B prompts the phonetics as "the month you want to inquire is X month", then there is a correlation between the request input node a and the node B.
The method for determining the correlation includes a pearson correlation coefficient method, a spearman correlation coefficient method, a kender correlation coefficient method, and the like, and this is not particularly limited in the embodiment of the present invention.
The preset knowledge graph can be established according to the nodes of different node types and the correlation among the node attributes of the nodes of different node types. Entities in the preset knowledge graph are nodes of different node types, and the relationship among the entities is the correlation among node attributes.
The node type and the node attribute of each candidate node and the node type and the node attribute of the current node are input into a preset knowledge graph, and the correlation between the node attribute of each candidate node and the node attribute of the current node can be obtained. And taking the candidate node with the highest correlation as the next node in the outbound flow to be edited.
According to the outbound flow editing method provided by the embodiment of the invention, the correlation between the node attribute of each candidate node and the node attribute of the current node is determined according to the preset knowledge graph, so that the next node is determined, the outbound flow editing efficiency is improved, the obtained outbound flow has strong correlation of each node, and the practicability is strong.
Based on the above embodiment, step 130 then includes:
determining a precision enhancement sample based on the node type and the node attribute of the next node and the node type and the node attribute of the current node;
and carrying out optimization training on the node prediction model based on the precision enhancement sample.
Specifically, the node prediction model may be optimally trained by using the next node and the current node to improve the accuracy of the node prediction model.
The node type and node attribute of the next node, and the node type and node attribute of the current node may be combined into one precision enhancement sample. And collecting a plurality of precision enhancement samples and constructing a precision enhancement sample set.
Because the precision enhancement sample comprises two nodes, and the latter node is selected by service personnel according to actual conditions, the reliability is high. The precision enhancement samples are input into the node prediction model for training, so that the prediction capability of the node prediction model on the node type of the next node can be further enhanced.
According to the outbound flow editing method provided by the embodiment of the invention, the determined next node and the current node are used as precision enhancement samples to train the node prediction model, so that the prediction accuracy of the node prediction model is improved, and the output result of the node prediction model has strong practicability and high accuracy.
Based on any of the above embodiments, the service scenario corresponding to the sample outbound flow and/or the sample node prompt phonetics is the same as the outbound flow to be edited.
Specifically, the service scenario may be a service query scenario, a service transaction scenario, a service marketing scenario, and the like. For example, the service query scenario includes a telephone charge query scenario, a balance query scenario, a tariff query scenario, and the like. Therefore, when the node prediction model is trained, the sample outbound flow and/or the sample node prompt phonetics of the same service scene can be selected according to the service scene of the outbound flow to be edited.
For example, when the outbound flow to be edited is a call charge query outbound flow, the sample outbound flow for training the node prediction model should select a sample outbound flow in a call charge query scenario, and should not select a sample outbound flow in a financial marketing scenario.
According to the outbound flow editing method provided by the embodiment of the invention, the sample outbound flow and/or the sample node prompt phonetics in the same service scene as the outbound flow to be edited are selected, and the node prediction model is trained, so that the prediction accuracy of the node prediction model is improved, the output result of the node prediction model is strong in practicability and high in accuracy.
Based on any embodiment, the node type comprises at least one of a request input node, a desired node, a judgment node and an alert tone node; the node attribute also comprises a node name and a dialect type;
the request input node is used for sending input prompt information to the client and determining a corresponding slot position based on the input prompt information, wherein the slot position is used for storing the input information of the client;
the expected node is used for acquiring input information of the client based on the slot position;
the judgment node is used for determining the selection result of the customer based on the input information of the previous node and a preset given value;
the prompt tone node is used for sending prompt information to the client based on the set prompt type, and the prompt type comprises text and/or voice.
Specifically, the node types in the outbound flow include a request input node, a desired node, a judgment node and an alert tone node. For each type of node, its node attributes include, in addition to the node prompt phonetics, the node name and the phonetics type. The phonetics type is the representation form of prompting phonetics, and comprises voice and text. The connection position of each node in the outbound flow can be changed by adopting a dragging mode, so that the outbound flow is edited.
And the node prediction model predicts the node type of the next node of the current node according to the association degree between the nodes of different node types obtained by learning from the sample outbound flow. For example, the degree of association between nodes of different node types may be embodied as a correlation coefficient between the nodes. The node prediction model learns that the correlation coefficient between the expected node and the request input node is 90%, the correlation coefficient between the expected node and the judgment node is 10%, and the correlation coefficient between the expected node and the prompt tone node is 50% from the sample outbound flow, so that the correlation degree between the expected node and the request input node is the highest. When the input current node is the expected node, the next node predicted by the node prediction model according to the current node can be the request input node.
For explaining each node, the following definitions are defined for the user and the client in the embodiment of the present invention: the user is an editor of the outbound flow, for example, a service person who edits the call charge to inquire the outbound flow. The client is an outbound call object, for example, a mobile phone holder who inquires own mobile phone call charge in a call charge inquiry scene.
The request input node is used for sending input prompt information to the client and determining a corresponding slot position based on the input prompt information, wherein the slot position is used for storing the input information of the client. For example, a request input node is used to obtain input from a customer. When editing the outbound flow, the user can drag the request input node to the main page to create the node which the user requests to input. The node attributes of the request input node further include: and the node name can be set by a user in a self-defined way. Type of dialog: similar to the normal mode of the alert tone, text and/or speech may be selected. And triggering a condition, wherein the input of the user is necessary input, and if not, the process can be ignored to continue to execute. The slot matching is used for creating input of a variable storage client, and the variable types such as month type, amount type and the like can be selected according to the information types to be input by the client. And (4) slot extraction, namely extracting information in the client input according to the slot defined by the user.
The desired node is used to obtain the input information of the client based on the slot. For example, the node is expected to be used to obtain data additionally provided by the customer. When the outbound flow is edited, the user can drag the page to create an expected node for slot extraction and variable selection. The node attributes of the desired node further include: and the node name can be set by a user in a self-defined way. And the variable type comprises slot extraction and variable selection. And the slot extraction is used for extracting the slot according to the input of the client. And selecting variables, namely selecting the existing variables of the current flow.
The judgment node is used for determining the selection result of the customer based on the input information of the previous node and the preset given value. For example, the decision node is used to make a branch decision. When editing the outbound flow, the user may drag to the main page, creating a node of the logical selection branch. Judging the node attribute of the node further comprises: and the node name can be set by a user in a self-defined way. And judging conditions configured in the subsequent connecting line attributes of the judging nodes. AND configuring judgment conditions of variables AND given values, wherein more than one judgment condition can be connected through logical operators AND AND OR.
The prompt tone node is used for sending prompt information to the client based on the set prompt type, and the prompt type comprises text and/or voice. For example, an alert tone node is used to play an alert tone typically used to prompt a user for input or to inform a piece of information. When the outbound flow is edited, the user can drag the prompt tone node to the main page to create a prompt tone broadcast node. The node attributes of the alert tone node further include: and the node name can be set by a user in a self-defined way. And selecting a mode, and selecting a normal mode and an advanced mode. And in a normal mode, the selectable prompt type is text or voice. Advanced mode: is a mode that can support Lua editing. And the text, the custom text and the known variables can be inserted to carry out voice synthesis broadcasting. And finally clicking and adding to generate text broadcast. Voice, broadcast for known recorded files. An existing voice file needs to be added.
Based on any of the above embodiments, the initial model of the node prediction model is a convolutional neural network or a cyclic neural network.
Specifically, the initial model may select a Convolutional Neural Network (CNN), a cyclic Neural Network (RNN), and the like, and the selection of the initial model is not specifically limited in the embodiment of the present invention.
Based on any of the above embodiments, the sample outbound flow includes a plurality of question-answer pairs sequentially connected in conversation order.
In particular, the data structure of the sample outbound flow may be in the form of a combination of multiple question-answer pairs. The question-answer pairs are connected in sequence according to the conversation sequence. The question in the question-answer pair may be taken as a node in the sample outbound flow.
According to the outbound flow editing method provided by the embodiment of the invention, the data structure of the sample outbound flow adopts a question-answer pair form, and the data structure is simple and convenient and is easy to obtain.
Based on any of the above embodiments, fig. 2 is a second schematic flow chart of the outbound flow editing method provided by the present invention, as shown in fig. 2, the method includes:
step one, obtaining a current node
The system gives the known conditions of current node type, attribute, dialect (intention dialect/request input dialect/prompt dialect) to the prediction model. For example: the current node is the start node and the intended terminology is "query charges".
Step two, a prediction model
And training a prediction model by combining a large amount of dialogue data and a large amount of dialogue process data to predict process nodes.
Step three, predicting the result
And giving a system prediction result through the prediction score and the comprehensive decision. The next node is predicted to be: 1. request input node 2, desired node 3, and judgment node
Step four, user selection
The user selects any one according to actual conditions. For example: "1, request input node".
Step five, iteration improvement
And storing the current node information and the predicted user selection node information into the model, thereby further improving the accuracy of the model.
Based on any of the above embodiments, fig. 3 is a schematic structural diagram of an outbound flow editing apparatus provided by the present invention, and as shown in fig. 3, the apparatus includes:
an obtaining unit 310, configured to obtain a node type and a node attribute of a current node in an outbound flow to be edited;
the prediction unit 320 is configured to input the node type and the node attribute of the current node into the node prediction model, and obtain a candidate node of a next node output by the node prediction model;
a determining unit 330, configured to determine a next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node;
the node attributes comprise node prompt phonetics, and the node prediction model is obtained based on a sample outbound flow and/or sample node prompt phonetics training.
The outbound flow editing device provided by the embodiment of the invention inputs the node type and the node attribute of the current node in the outbound flow to be edited into the node prediction model to obtain a plurality of candidate nodes of the next node, so as to determine the next node corresponding to the current node, wherein the node prediction model is obtained based on the sample outbound flow and sample node prompt phonetics training, can accurately predict the node in the outbound flow to be edited and the prompt phonetics of the node, guides a service worker to carry out outbound flow editing, reduces the difficulty of the outbound flow editing, improves the efficiency of the outbound flow editing, and has strong relevance and practicability of each node in the obtained outbound flow.
Based on any embodiment above, the device includes:
the training unit is used for determining a plurality of sample outbound flows and a plurality of sample node prompt phonetics;
determining the association degree between nodes of different node types in each sample outbound flow and the evaluation label of each sample node prompt phonetics; the evaluation label is used for representing the semantic expression capability of the sample node prompt phonetics;
and training the initial model based on the correlation degree between the nodes of different node types in the sample outbound flows and each sample outbound flow and the evaluation labels of the prompt phonetics of the sample nodes and each sample node to obtain a node prediction model.
Based on any of the above embodiments, the determining unit 330 is configured to:
determining client intention prediction information based on the node type and/or node attribute of the current node;
and determining the similarity between the client intention prediction information and the node prompt phonetics in the node attributes of the candidate nodes, and taking the candidate node with the highest similarity as the next node in the outbound flow to be edited.
Based on any of the above embodiments, the determining unit 330 is configured to:
determining the prediction weight of each node type based on the number of nodes of each node type in the sample outbound flow and the total number of nodes in the sample outbound flow;
and determining the next node in the outbound flow to be edited based on the prediction weight of each node type and the node type of the candidate node.
Based on any of the above embodiments, the determining unit 330 is configured to:
determining the correlation between the node attribute of each candidate node and the node attribute of the current node based on the node type and the node attribute of each candidate node, the node type and the node attribute of the current node and a preset knowledge graph;
determining a next node in the outbound flow to be edited based on the correlation between the node attribute of each candidate node and the node attribute of the current node;
wherein the preset knowledge graph is determined based on the nodes of different node types and the correlation between the node attributes of the nodes of different node types.
Based on any embodiment above, the apparatus further comprises:
the precision enhancing unit is used for determining a precision enhancing sample based on the node type and the node attribute of the next node and the node type and the node attribute of the current node; and training the node prediction model based on the precision enhancement sample.
Based on any of the above embodiments, the service scenario corresponding to the sample outbound flow and/or the sample node prompt phonetics is the same as the outbound flow to be edited.
Based on any embodiment, the node type comprises at least one of a request input node, a desired node, a judgment node and an alert tone node; the node attribute also comprises a node name and a dialect type;
the request input node is used for sending input prompt information to the client and determining a corresponding slot position based on the input prompt information, wherein the slot position is used for storing the input information of the client;
the expected node is used for acquiring input information of the client based on the slot position;
the judgment node is used for determining the selection result of the customer based on the input information of the previous node and a preset given value;
the prompt tone node is used for sending prompt information to the client based on the set prompt type, and the prompt type comprises text and/or voice.
Based on any of the above embodiments, the initial model of the node prediction model is a convolutional neural network or a cyclic neural network.
Based on any of the above embodiments, the sample outbound flow includes a plurality of question-answer pairs sequentially connected in conversation order.
Based on any of the above embodiments, fig. 4 is a schematic structural diagram of an electronic device provided by the present invention, and as shown in fig. 4, the electronic device may include: a Processor (Processor)410, a communication Interface (communication Interface)420, a Memory (Memory)430 and a communication Bus (communication Bus)440, wherein the Processor 410, the communication Interface 420 and the Memory 430 are communicated with each other via the communication Bus 440. The processor 410 may call logical commands in the memory 430 to perform the following method:
acquiring the node type and node attribute of a current node in an outbound flow to be edited;
inputting the node type and the node attribute of the current node into a node prediction model to obtain a candidate node of a next node output by the node prediction model;
determining a next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node;
the node attributes comprise node prompt phonetics, and the node prediction model is obtained based on a sample outbound flow and/or sample node prompt phonetics training.
In addition, the logic commands in the memory 430 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic commands are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of commands for enabling a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The processor in the electronic device provided in the embodiment of the present invention may call a logic instruction in the memory to implement the method, and the specific implementation manner of the method is consistent with the implementation manner of the method, and the same beneficial effects may be achieved, which is not described herein again.
Embodiments of the present invention further provide a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented to perform the method provided in the foregoing embodiments when executed by a processor, and the method includes:
acquiring the node type and node attribute of a current node in an outbound flow to be edited;
inputting the node type and the node attribute of the current node into a node prediction model to obtain a candidate node of a next node output by the node prediction model;
determining a next node in the outbound flow to be edited based on the node type and the node attribute of the candidate node;
the node attributes comprise node prompt phonetics, and the node prediction model is obtained based on a sample outbound flow and/or sample node prompt phonetics training.
When the computer program stored on the non-transitory computer readable storage medium provided in the embodiments of the present invention is executed, the method is implemented, and the specific implementation manner of the method is consistent with the implementation manner of the method, and the same beneficial effects can be achieved, which is not described herein again.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes commands for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110636513.1A CN113570324A (en) | 2021-06-08 | 2021-06-08 | Outbound call flow editing method, device, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110636513.1A CN113570324A (en) | 2021-06-08 | 2021-06-08 | Outbound call flow editing method, device, electronic device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113570324A true CN113570324A (en) | 2021-10-29 |
Family
ID=78161745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110636513.1A Pending CN113570324A (en) | 2021-06-08 | 2021-06-08 | Outbound call flow editing method, device, electronic device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113570324A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114466106A (en) * | 2021-12-23 | 2022-05-10 | 北京声智科技有限公司 | Test data generation method, device, equipment and medium of outbound system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1107555A2 (en) * | 1999-12-10 | 2001-06-13 | Genesys Telecommunications Laboratories, Inc. | Integrated communication center functionality for WAP devices |
| CN110457709A (en) * | 2019-08-16 | 2019-11-15 | 北京一链数云科技有限公司 | Outgoing call dialog process method, apparatus and server |
| CN111159501A (en) * | 2019-11-22 | 2020-05-15 | 杭州蛋壳商务信息技术有限公司 | Method for establishing passenger judging model based on multilayer neural network and passenger judging method |
| CN112185358A (en) * | 2020-08-24 | 2021-01-05 | 维知科技张家口有限责任公司 | Intention recognition method, model training method, device, equipment and medium |
| CN112202978A (en) * | 2020-08-24 | 2021-01-08 | 维知科技张家口有限责任公司 | Intelligent outbound call system, method, computer system and storage medium |
-
2021
- 2021-06-08 CN CN202110636513.1A patent/CN113570324A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1107555A2 (en) * | 1999-12-10 | 2001-06-13 | Genesys Telecommunications Laboratories, Inc. | Integrated communication center functionality for WAP devices |
| CN110457709A (en) * | 2019-08-16 | 2019-11-15 | 北京一链数云科技有限公司 | Outgoing call dialog process method, apparatus and server |
| CN111159501A (en) * | 2019-11-22 | 2020-05-15 | 杭州蛋壳商务信息技术有限公司 | Method for establishing passenger judging model based on multilayer neural network and passenger judging method |
| CN112185358A (en) * | 2020-08-24 | 2021-01-05 | 维知科技张家口有限责任公司 | Intention recognition method, model training method, device, equipment and medium |
| CN112202978A (en) * | 2020-08-24 | 2021-01-08 | 维知科技张家口有限责任公司 | Intelligent outbound call system, method, computer system and storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114466106A (en) * | 2021-12-23 | 2022-05-10 | 北京声智科技有限公司 | Test data generation method, device, equipment and medium of outbound system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7635369B2 (en) | Systems and methods for bot authoring by mining intent from natural language conversations | |
| US11425064B2 (en) | Customized message suggestion with user embedding vectors | |
| US11315569B1 (en) | Transcription and analysis of meeting recordings | |
| JP2023545947A6 (en) | Systems and methods for bot authoring by mining intent from conversation data via intent seeding | |
| JP2023543032A6 (en) | Systems and methods for bot authoring by mining intent from natural language conversations | |
| CN104598445B (en) | Automatically request-answering system and method | |
| US20180157960A1 (en) | Scalable curation system | |
| CN110825858A (en) | An Intelligent Interactive Robot System Applied in Customer Service Center | |
| CN112948534A (en) | Interaction method and system for intelligent man-machine conversation and electronic equipment | |
| CN111951782A (en) | Voice question answering method and apparatus, computer readable storage medium and electronic device | |
| US12158902B2 (en) | Transition-driven transcript search | |
| CN117453895B (en) | Intelligent customer service response method, device, equipment and readable storage medium | |
| KR20200139599A (en) | Method and appratus for providing counseling service | |
| CN117312521A (en) | Processing method for intelligent customer service dialogue and related products | |
| CN119669405A (en) | A customer service processing method, device, equipment, storage medium and product based on big data | |
| CN117059095A (en) | IVR-based service providing method and device, computer equipment and storage medium | |
| US20220207066A1 (en) | System and method for self-generated entity-specific bot | |
| CN113570324A (en) | Outbound call flow editing method, device, electronic device and storage medium | |
| US12437159B2 (en) | Methods and systems for enhanced searching of conversation data and related analytics in a contact center | |
| CN118395227A (en) | Social assistance method, device, electronic device and computer-readable storage medium | |
| JP7169031B1 (en) | Program, information processing device, information processing system, information processing method, information processing terminal | |
| US20230259541A1 (en) | Intelligent Assistant System for Conversational Job Search | |
| CN110929005A (en) | Task follow-up method, device, device and storage medium based on sentiment analysis | |
| JP7169030B1 (en) | Program, information processing device, information processing system, information processing method, information processing terminal | |
| CN118283182A (en) | AI-based intelligent voice call prediction method, program product, device and medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211029 |