CN114581162B - Method and device for predicting order in customer service conversation process and electronic equipment - Google Patents
Method and device for predicting order in customer service conversation process and electronic equipment Download PDFInfo
- Publication number
- CN114581162B CN114581162B CN202210495977.XA CN202210495977A CN114581162B CN 114581162 B CN114581162 B CN 114581162B CN 202210495977 A CN202210495977 A CN 202210495977A CN 114581162 B CN114581162 B CN 114581162B
- Authority
- CN
- China
- Prior art keywords
- dialogue data
- information
- unit
- data
- neural network
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method and a device for predicting a service formation in a customer service conversation process and electronic equipment, wherein the method comprises the following steps: acquiring historical dialogue data; inquiring the order log, and determining a buyer order log within a set time range; classifying and labeling historical dialogue data; extracting characteristic parameters of historical dialogue data and processing the characteristic parameters; establishing a neural network model; training the neural network model to obtain a one-generation prediction model; collecting current dialogue data and extracting characteristic parameters of the current dialogue data; inputting characteristic parameters of current dialogue data; loading a one-generation prediction model, and processing the characteristic parameters by the one-generation prediction model; and outputting the prediction result of the single rate. The method obtains the single-forming prediction model through the neural network model training, and predicts the single-forming rate in real time in conversation chatting by using the single-forming prediction model, thereby being beneficial to customer service staff to adjust conversation behaviors in real time.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a method and a device for predicting a service completion in a customer service conversation process and electronic equipment.
Background
The e-commerce sale is higher in the sale link, and in the sale process, the customer service answers various questions for the buyer. However, due to the difference of the capacities of the customer service staff, the final ordering result of the same buyer is influenced by the capacities of the customer service staff. By collecting the single rate data and summarizing the single rate data, the sales experience of customer service staff can be enriched; in the prior art, the statistics of the success rate can be given in a posterior mode after a plurality of groups of data are accumulated, the success rate cannot be predicted in real time in conversation chat, and the real-time conversation behavior adjustment of customer service personnel is not facilitated. Traditionally, a single sentence is subjected to sentence vector coding, and then other operations such as semantic recognition, emotion recognition and the like are carried out, and a plurality of sentences of one conversation are not used together to predict a single rate.
Disclosure of Invention
In order to solve the technical problem, the invention provides a method and a device for predicting a service formation in a customer service conversation process and electronic equipment.
The embodiment of the invention is realized by the following technical scheme:
in a first aspect, the present disclosure provides a method for predicting a service completion in a customer service conversation process, including the following steps:
acquiring historical dialogue data;
inquiring an order log according to the historical dialogue data, and determining a buyer order log corresponding to the historical dialogue data within a set time range;
according to the buyer order log, carrying out classification and labeling on the historical dialogue data according to the result of the final order in the set time range;
extracting characteristic parameters of the historical dialogue data according to the historical dialogue data and the classification and labeling result, and processing the characteristic parameters of the historical dialogue data;
establishing a neural network model according to the processed characteristic parameters of the historical dialogue data: determining feature parameters of the processed historical dialogue data according to the feature parameters of the processed historical dialogue data, wherein the feature parameters comprise word information contained in statement information, role information corresponding to the statement information and time information corresponding to the statement information; encoding word information contained in the statement information, role information corresponding to the statement information and time information corresponding to the statement information to determine a statement vector of the statement information; coding the sentence vector of the statement information to determine a dialogue vector; secondly, classifying the dialogue vectors by utilizing a softmax function, and determining an output result of the neural network model; calculating a cross entropy loss value by using a cross entropy loss function; determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss;
training the neural network model by using the historical dialogue data and the classification and labeling result to obtain a one-dimensional prediction model;
collecting current dialogue data and extracting characteristic parameters of the current dialogue data;
inputting characteristic parameters of the current dialogue data into the simple prediction model;
loading the one-generation prediction model, and processing the characteristic parameters of the current dialogue data by the one-generation prediction model;
and the list forming prediction model outputs the prediction result of the list forming rate of the current dialogue data.
In a second aspect, the present disclosure provides a service-to-order prediction apparatus in a customer service conversation process, including an obtaining unit, a query unit, a classification labeling unit, a first feature extraction unit, a neural network model establishing unit, a neural network model training unit, a second feature extraction unit, an input unit, a processing unit, and an output unit:
the acquisition unit is used for acquiring historical dialogue data;
the inquiry unit is used for inquiring an order log according to the historical dialogue data and determining a buyer order log corresponding to the historical dialogue data within a set time range;
the classification and labeling unit is used for classifying and labeling the historical dialogue data according to the formed order result in the set time range according to the buyer order log;
the first feature extraction unit is used for extracting feature parameters of the historical dialogue data according to the historical dialogue data and the classification labeling result and processing the feature parameters of the historical dialogue data;
the neural network model establishing unit is used for establishing a neural network model according to the processed characteristic parameters of the historical dialogue data;
the neural network model building unit includes: the device comprises a first determining unit, a first encoding unit, a second encoding unit, a classifying unit, a first calculation processing unit and a second determining unit;
the first determining unit is configured to determine feature parameters of the processed historical dialogue data according to the feature parameters of the processed historical dialogue data, where the feature parameters include word information included in statement information, role information corresponding to the statement information, and time information corresponding to the statement information;
the first encoding unit is configured to encode word information included in the statement information, role information corresponding to the statement information, and time information corresponding to the statement information, and determine a statement vector of the statement information;
the second encoding unit is configured to encode a sentence vector of the sentence information and determine a dialogue vector;
the classification unit is used for carrying out secondary classification on the dialogue vectors by utilizing a softmax function and determining an output result of the neural network model;
the first calculation processing unit is used for calculating the value of cross entropy loss by using a cross entropy loss function;
the second determining unit is used for determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss;
the neural network model training unit is used for training the neural network model by using the historical dialogue data and the classification and labeling result to obtain a one-dimensional prediction model;
the second feature extraction unit is used for collecting current dialogue data and extracting feature parameters of the current dialogue data;
the input unit is used for inputting the characteristic parameters of the current dialogue data to the single prediction model;
the processing unit is used for loading the one-generation prediction model, and the one-generation prediction model is used for processing the characteristic parameters of the current dialogue data;
the output unit is used for outputting the prediction result of the single forming rate of the current dialogue data by the single forming prediction model.
In a third aspect, the present disclosure provides an electronic device, comprising:
a processor and a memory;
the memory is used for storing computer operation instructions;
the processor is used for executing the one-to-one prediction method in the customer service conversation process by calling the computer operation instruction.
The beneficial effects of the invention are: the method obtains the single-forming prediction model through the training of the neural network model, and predicts the single-forming rate in real time in conversation chatting by using the single-forming prediction model, thereby being beneficial to customer service staff to adjust conversation behaviors in real time.
On the basis of the technical scheme, the invention can be further improved as follows.
Further, the establishing a neural network model according to the processed feature parameters of the historical dialogue data includes:
determining characteristic parameters of the processed historical dialogue data according to the characteristic parameters of the processed historical dialogue data, wherein the characteristic parameters comprise word information contained in statement information, role information corresponding to the statement information and time information corresponding to the statement information;
encoding word information contained in the statement information, role information corresponding to the statement information and time information corresponding to the statement information, and determining a statement vector of the statement information;
coding the sentence vector of the sentence information to determine a dialogue vector;
secondly, classifying the dialogue vectors by utilizing a softmax function, and determining an output result of the neural network model;
calculating a cross entropy loss value by using a cross entropy loss function;
and determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss.
The method has the advantages that the historical dialogue data is coded twice through sentence vectors and dialogue vectors, so that discrete character information is continuous, similar dialogue or scenes or flows are found through the continuous vectors, and the generalization capability of the neural network model is improved; according to the value of the cross entropy loss, the loss is minimized, model parameters can be adjusted through an iteration mode according to the predicted value, the result of real data is approached step by step, and the accuracy of the output result of the neural network model is improved.
Further, the acquiring historical dialogue data further includes: and filtering the historical dialogue data.
The beneficial effect of adopting the further scheme is that the historical dialogue data without the characteristic parameters can be filtered by filtering the historical dialogue data.
Further, the characteristic parameters comprise statement information, role information and time information; the statement information comprises conversation content; the role information comprises the identity of a dialog person; the time information comprises time data corresponding to the statement information and time interval data between the current conversation and the previous conversation of the current conversation
The beneficial effect of adopting the further scheme is that the characteristic parameters such as statement information, role information and time information can better reflect the influence of the conversation on the single result.
Further, the processing the characteristic parameters of the historical dialogue data includes: carrying out natural language processing on the statement information by using a natural language model, and coding the statement information; encoding the role information by using a one-hot encoding mode; and coding the time data corresponding to the statement information by using a one-hot coding mode, and segmenting and coding the time interval data between the current conversation and the previous adjacent conversation of the current conversation by using the one-hot coding mode.
The beneficial effect of adopting the further scheme is that the coding of the characteristic parameters is realized.
Further, the generating unit prediction model outputs a prediction result of the generating unit rate of the current dialogue data, and further includes:
calculating the change rate of the single forming rate of the current dialogue data according to the prediction result of the single forming rate of the current dialogue data;
and displaying the prediction result of the single rate of the current dialogue data and the change rate of the single rate of the current dialogue data.
The method has the advantages that the display of the single-rate prediction result and the change rate is realized, and the method is favorable for customer service staff to adjust conversation behaviors in real time.
Drawings
Fig. 1 is a flowchart of a policy prediction method in a customer service session process according to embodiment 1 of the present invention;
fig. 2 is a schematic diagram illustrating a principle of establishing a neural network model in embodiment 1 of the present invention;
fig. 3 is a schematic diagram of a service-order prediction device in a customer service session process according to embodiment 2 of the present invention;
fig. 4 is a schematic diagram of an electronic device according to embodiment 3 of the present invention.
Icon: 201-word information; 202-role information corresponding to the statement information; 203-time information corresponding to the statement information; 204-sentence vectors of word information; 205-sentence vectors of sentence information; 206-softmax layer; 207-class II; 40-an electronic device; 410-a processor; 420-a bus; 430-a memory; 440-a transceiver.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Example 1
As an embodiment, as shown in fig. 1, to solve the above technical problem, the embodiment provides a method for predicting a customer service form in a customer service conversation process, which includes the following steps:
acquiring historical dialogue data;
inquiring an order log according to the historical dialogue data, and determining a buyer order log corresponding to the historical dialogue data within a set time range;
according to the buyer order log, carrying out classification and labeling on the historical dialogue data according to the result of the final order in the set time range;
extracting characteristic parameters of the historical dialogue data according to the historical dialogue data and the classification labeling result, and processing the characteristic parameters of the historical dialogue data;
establishing a neural network model according to the processed characteristic parameters of the historical dialogue data: determining feature parameters of the processed historical dialogue data according to the feature parameters of the processed historical dialogue data, wherein the feature parameters comprise word information contained in statement information, role information corresponding to the statement information and time information corresponding to the statement information; encoding word information contained in the statement information, role information corresponding to the statement information and time information corresponding to the statement information to determine a statement vector of the statement information; coding the sentence vector of the sentence information to determine a dialogue vector; secondly, classifying the dialogue vectors by utilizing a softmax function, and determining an output result of the neural network model; calculating a cross entropy loss value by using a cross entropy loss function; determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss;
training the neural network model by using the historical dialogue data and the classification and labeling result to obtain a one-dimensional prediction model;
collecting current dialogue data and extracting characteristic parameters of the current dialogue data;
inputting characteristic parameters of the current dialogue data into the single prediction model;
loading the one-to-one prediction model, wherein the one-to-one prediction model processes the characteristic parameters of the current dialogue data;
and the single-forming prediction model outputs the prediction result of the single-forming rate of the current dialogue data.
Optionally, as shown in fig. 2, the establishing a neural network model according to the processed feature parameters of the historical dialogue data includes:
determining feature parameters of the processed historical dialogue data according to the feature parameters of the processed historical dialogue data, wherein the feature parameters comprise word information contained in statement information 201, role information 202 corresponding to the statement information 201, and time information 203 corresponding to the statement information 201;
encoding word information included in the term information 201, role information 202 corresponding to the term information 201, and time information 203 corresponding to the term information 201, and determining a term vector 205 of the term information 201;
encoding a sentence vector 205 of the sentence information to determine a dialogue vector 206; in fig. 2, 201 represents word information, 202 represents role information corresponding to the word information 201, 203 represents time information corresponding to the word information 201, 204 represents a word vector of the word information, and 205 represents a word vector of the word information 201, and the word vector of the word information 201 includes the word vector 204 of the word information 201, the word vector of the role information 202 corresponding to the word information 201, and the word vector of the time information 203 corresponding to the word information 201.
And classifying 207 the dialogue vectors by utilizing a softmax layer 206, and determining an output result of the neural network model.
Calculating a cross entropy loss value by using a cross entropy loss function;
and determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss.
In the actual application process, optionally, the number N of sentences of the dialog is determined according to the historical dialog data. For a dialog with the actual sentence number larger than a set value N (such as N = 50), taking the reciprocal N sentences as sample data for training the neural network model, and for a dialog with less than N sentences, the requirement of the model input length can be met by adopting a mode of complementing 0 vectors. Optionally, the sentence vector of the statement information, the role information corresponding to the statement information, and the time information corresponding to the statement information are encoded by using an embedding algorithm + cnn/lstm/transformer/bert algorithm, and the sentence vector of the statement information, the encoded role characteristics, and the encoded time characteristics are combined to form the sentence vector of the statement information.
Wherein, the sentence vector is the vector representation of the sentence; a dialog vector is a vector representation of dialog information.
Optionally, the dialog vector is subjected to secondary classification by using a softmax function, an output result of the neural network model is determined, and the output result is two result tags and a probability corresponding to each result.
In addition, the cross entropy loss value is calculated, the parameters of the neural network model are obtained by minimizing the cross entropy loss, the smaller the cross entropy loss is, the more accurate the parameters of the neural network model are, and finally the single prediction model is obtained, so that the model parameters can be adjusted, and the accuracy of the output result of the neural network model can be improved.
Optionally, the acquiring historical dialogue data further includes: and filtering the historical dialogue data according to the characteristic parameters of the historical dialogue data.
In the practical application process, by filtering the historical dialogue data, the historical dialogue data without characteristic parameters, such as a dialogue in which a buyer does not send a message, an after-sale dialogue, a dialogue in which a customer serves a recommended advertisement sent for an old user, and the like, can be filtered.
Optionally, the feature parameters include statement information, role information, and time information; the statement information comprises conversation content; the role information comprises the identity of a dialog person; the time information comprises time data corresponding to the statement information and a time interval between the current conversation and the previous conversation of the current conversation, wherein the time interval is a time interval between adjacent conversations.
In the practical application process, the characteristic parameters such as statement information, role information and time information can better reflect the influence of the conversation on the single result.
Optionally, the processing the characteristic parameters of the historical dialogue data includes: carrying out natural language processing on the statement information by using a natural language model, and coding the statement information; encoding the role information by using a one-hot encoding mode; and coding the time data corresponding to the statement information by using a one-hot coding mode, and segmenting and coding the time interval data between the current conversation and the previous adjacent conversation of the current conversation by using the one-hot coding mode.
In the actual application process, the role information is encoded by using a one-hot encoding method, for example: coding buyer role information to obtain (1, 0, 0), coding seller role information to obtain (0,1, 0), and coding robot role information to obtain (0, 0, 1); and the time interval data between the current conversation and the previous adjacent conversation of the current conversation is segmented and coded by using a one-hot coding mode, and the method comprises the following steps: segmenting the time interval data between the current dialog and the previous dialog of the current dialog according to the time interval size, determining the number of segmentation segments, and obtaining an encoding result according to the position of the time interval data size in the segmentation result, for example: the values of the time intervals are: 1, 2, 3.; segmenting the value of the time interval to obtain: 1-10, 11-20, 21-30, 31-60,61-120,121-300,301-600, which is greater than 600, and 8 segments in total, wherein according to the bit number of the value of the time interval in the segment, the bit number data is 1, the rest bit data is 0, if the time interval is 25 seconds, 25 is in the third segment, the corresponding third bit data is 1, and the rest bit data is 0, and the obtained encoded vector is: (0,0,1,0,0,0,0,0).
The characteristic parameters are processed in a coding mode, the statement information is coded in a token mode, potential relations between semantics, similar scenes and the like can be extracted, and the generalization of a neural network model is easier.
Optionally, the generating single prediction model outputs a prediction result of the generating single rate of the current dialogue data, and further includes:
calculating the change rate of the single forming rate of the current dialogue data according to the prediction result of the single forming rate of the current dialogue data;
and displaying the prediction result of the single rate of the current dialogue data and the change rate of the single rate of the current dialogue data.
In the actual application process, optionally, the prediction service is started through the prediction service interface, and the neural network model is loaded to obtain the prediction service; and starting the prediction service by acquiring a post request because the data of the conversation is more.
The prediction service is divided into two interfaces, one interface only returns the real-time single-rate prediction value of the current conversation, and the other interface returns a real-time single-rate prediction value according to the information when each conversation is finished after the conversation is received.
Optionally, the predicted service interface includes two interfaces: one method is to obtain a predicted value of the single-forming rate in real time according to the conversation content of the current sentence; and the other method is that after the conversation is obtained, a predicted value of the real-time single-forming rate is obtained according to all conversation contents of the current conversation or a set number of conversation contents when each sentence of conversation is ended.
The display of the single-rate prediction result and the change rate is realized, and the real-time adjustment of the conversation behavior by customer service personnel is facilitated.
Optionally, when the current session data is collected, the current session content is acquired in real time.
Optionally, when the current real-time single rate is displayed, the current real-time single rate is compared with the single rate obtained by the previous dialogue, and the comparison result such as rising and falling is distinguished by different colors, so that the reminding effect is achieved.
Optionally, after the customer service staff inputs the conversation content through the prediction service interface, the sent real-time single rate is displayed, so that the customer service staff can try to reply different conversation contents, and the real-time conversation behavior adjustment of the customer service staff is facilitated.
Example 2
Based on the same principle as the method shown in embodiment 1 of the present invention, as shown in fig. 3, an embodiment of the present invention further provides a device for predicting a customer service form in a customer service conversation process, which includes an obtaining unit, a query unit, a classification labeling unit, a first feature extraction unit, a neural network model establishing unit, a neural network model training unit, a second feature extraction unit, an input unit, a processing unit, and an output unit:
an acquisition unit configured to acquire historical dialogue data;
the query unit is used for querying the order log according to the historical conversation data and determining the buyer order log corresponding to the historical conversation data within a set time range;
the classification marking unit is used for classifying and marking the historical dialogue data according to the formed order result in the set time range according to the buyer order log;
the first feature extraction unit is used for extracting feature parameters of the historical dialogue data according to the historical dialogue data and the classification labeling result and processing the feature parameters of the historical dialogue data;
the neural network model establishing unit is used for establishing a neural network model according to the characteristic parameters of the processed historical dialogue data;
the neural network model training unit is used for training the neural network model by using historical dialogue data and a classification labeling result to obtain a one-to-one prediction model;
the second characteristic extraction unit is used for collecting the current dialogue data and extracting the characteristic parameters of the current dialogue data;
the input unit is used for inputting the characteristic parameters of the current dialogue data into the monotone prediction model;
the processing unit is used for loading a one-dimensional prediction model and processing the characteristic parameters of the current dialogue data by the one-dimensional prediction model;
and the output unit is used for outputting the prediction result of the single forming rate of the current dialogue data by the single forming prediction model.
Optionally, the neural network model building unit includes a first determining unit, a first encoding unit, a second encoding unit, a classifying unit, a first computing unit, and a second determining unit:
a first determining unit, configured to determine, according to feature parameters of the processed historical dialogue data, that the feature parameters of the processed historical dialogue data correspond to, where the feature parameters include word information included in statement information, role information corresponding to the statement information, and time information corresponding to the statement information;
a first encoding unit, configured to encode word information included in the sentence information, role information corresponding to the sentence information, and time information corresponding to the sentence information, and determine a sentence vector of the sentence information;
the second coding unit is used for coding the sentence vectors of the sentence information and determining dialogue vectors;
the classification unit is used for carrying out secondary classification on the dialogue vectors by utilizing a softmax function and determining an output result of the neural network model;
a first calculation processing unit for calculating a value of cross entropy loss using a cross entropy loss function;
and the second determining unit is used for determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss.
Optionally, the obtaining unit further includes a filtering unit, configured to filter the historical dialogue data.
Optionally, the characteristic parameters include statement information, role information, and time information; the statement information comprises conversation content; the role information comprises the identity of a dialog person; the time information comprises time data corresponding to the statement information and a time interval between the current conversation and the previous conversation of the current conversation, wherein the time interval is a time interval between adjacent conversations.
Optionally, the first feature extraction unit includes:
a natural language processing unit for performing natural language processing on the sentence information by using the natural language model;
an encoding unit for encoding the sentence information; encoding the character information by using a one-hot encoding mode; and coding time data corresponding to the sentence information by using a one-hot coding mode, and segmenting and coding time interval data between the current conversation and the conversation adjacent to the previous sentence of the current conversation by using the one-hot coding mode.
Optionally, the output unit further includes:
the second calculation processing unit is used for calculating the change rate of the single forming rate of the current dialogue data according to the prediction result of the single forming rate of the current dialogue data;
and the display unit is used for displaying the prediction result of the single forming rate of the current dialogue data and the change rate of the single forming rate of the current dialogue data.
Example 3
Based on the same principle as the method shown in the embodiment of the present invention, an embodiment of the present invention further provides an electronic device, as shown in fig. 4, which may include but is not limited to: a processor and a memory; a memory for storing a computer program; a processor for executing the method according to any of the embodiments of the present invention by calling the computer program.
In an alternative embodiment, an electronic device is provided, the electronic device 40 shown in fig. 4 comprising: a processor 410 and a memory 430. Wherein processor 410 is coupled to memory 430, such as via bus 420.
Optionally, the electronic device 40 may further include a transceiver 440, and the transceiver 440 may be used for data interaction between the electronic device and other electronic devices, such as transmission of data and/or reception of data. It should be noted that the transceiver 440 is not limited to one in practical applications, and the structure of the electronic device 40 is not limited to the embodiment of the present invention.
The Processor 410 may be a CPU (Central Processing Unit), a general-purpose Processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other Programmable logic device, a transistor logic device, a hardware component, or any combination thereof. Which may implement or perform the various illustrative logical blocks, modules, and circuits described in connection with the disclosure. The processor 410 may also be a combination of computing functions, e.g., comprising one or more microprocessors in combination, a DSP and a microprocessor in combination, or the like.
The Memory 430 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical Disc storage, optical Disc storage (including Compact Disc, laser Disc, optical Disc, digital versatile Disc, blu-ray Disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to these.
The memory 430 is used for storing application program codes (computer programs) for performing aspects of the present invention and is controlled in execution by the processor 410. The processor 410 is configured to execute application program code stored in the memory 430 to implement the aspects illustrated in the foregoing method embodiments.
The above is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made to the present invention by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A method for predicting a successful order in a customer service conversation process is characterized by comprising the following steps:
acquiring historical dialogue data;
inquiring an order log according to the historical dialogue data, and determining a buyer order log corresponding to the historical dialogue data within a set time range;
according to the buyer order log, carrying out classification and labeling on the historical dialogue data according to the result of the final order in the set time range;
extracting characteristic parameters of the historical dialogue data according to the historical dialogue data and the classification labeling result, and processing the characteristic parameters of the historical dialogue data;
establishing a neural network model according to the processed characteristic parameters of the historical dialogue data: determining feature parameters of the processed historical dialogue data according to the feature parameters of the processed historical dialogue data, wherein the feature parameters comprise word information contained in statement information, role information corresponding to the statement information and time information corresponding to the statement information; encoding word information contained in the statement information, role information corresponding to the statement information and time information corresponding to the statement information to determine a statement vector of the statement information; coding the sentence vector of the sentence information to determine a dialogue vector; secondly, classifying the dialogue vectors by utilizing a softmax function, and determining an output result of the neural network model; calculating a cross entropy loss value by using a cross entropy loss function; determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss;
training the neural network model by using the historical dialogue data and the classification and labeling result to obtain a one-dimensional prediction model;
collecting current dialogue data and extracting characteristic parameters of the current dialogue data;
inputting characteristic parameters of the current dialogue data into the simple prediction model;
loading the one-to-one prediction model, wherein the one-to-one prediction model processes the characteristic parameters of the current dialogue data;
and the single-forming prediction model outputs the prediction result of the single-forming rate of the current dialogue data.
2. The method of claim 1, wherein the obtaining historical dialogue data further comprises: and filtering the historical dialogue data.
3. The method for predicting the form of a customer service dialog process as claimed in claim 1, wherein the characteristic parameters comprise statement information, role information and time information; the statement information comprises conversation content; the role information comprises the identity of a dialog person; the time information comprises time data corresponding to the statement information and time interval data between the current conversation and the previous conversation of the current conversation.
4. The method of claim 3, wherein the processing the characteristic parameters of the historical dialogue data comprises: carrying out natural language processing on the statement information by using a natural language model, and coding the statement information; encoding the role information by using a one-hot encoding mode; and coding the time data corresponding to the statement information by using a one-hot coding mode, and segmenting and coding the time interval data between the current conversation and the previous conversation of the current conversation by using the one-hot coding mode.
5. The method of claim 1, wherein the uni-prediction model outputs the prediction result of the uni-rate of the current session data, and further comprises:
calculating the change rate of the single forming rate of the current dialogue data according to the prediction result of the single forming rate of the current dialogue data;
and displaying the prediction result of the single rate of the current dialogue data and the change rate of the single rate of the current dialogue data.
6. The forming order prediction device in the customer service conversation process is characterized by comprising an acquisition unit, a query unit, a classification labeling unit, a first feature extraction unit, a neural network model establishment unit, a neural network model training unit, a second feature extraction unit, an input unit, a processing unit and an output unit:
the acquisition unit is used for acquiring historical dialogue data;
the query unit is used for querying an order log according to the historical dialogue data and determining a buyer order log corresponding to the historical dialogue data within a set time range;
the classification and labeling unit is used for classifying and labeling the historical dialogue data according to the formed order result in the set time range according to the buyer order log;
the first feature extraction unit is used for extracting feature parameters of the historical dialogue data according to the historical dialogue data and the classification labeling result and processing the feature parameters of the historical dialogue data;
the neural network model establishing unit is used for establishing a neural network model according to the processed characteristic parameters of the historical dialogue data;
the neural network model building unit includes: the device comprises a first determining unit, a first encoding unit, a second encoding unit, a classifying unit, a first calculation processing unit and a second determining unit;
the first determining unit is configured to determine, according to feature parameters of the processed historical dialogue data, where the feature parameters include word information included in statement information, role information corresponding to the statement information, and time information corresponding to the statement information;
the first encoding unit is configured to encode word information included in the sentence information, role information corresponding to the sentence information, and time information corresponding to the sentence information, and determine a sentence vector of the sentence information;
the second encoding unit is configured to encode a sentence vector of the sentence information and determine a dialogue vector;
the classification unit is used for carrying out secondary classification on the dialogue vectors by utilizing a softmax function and determining an output result of the neural network model;
the first calculation processing unit is used for calculating the value of cross entropy loss by using a cross entropy loss function;
the second determining unit is used for determining the accuracy of the output result of the neural network model according to the value of the cross entropy loss;
the neural network model training unit is used for training the neural network model by using the historical dialogue data and the classification and labeling result to obtain a one-dimensional prediction model;
the second feature extraction unit is used for collecting current dialogue data and extracting feature parameters of the current dialogue data;
the input unit is used for inputting the characteristic parameters of the current dialogue data to the single prediction model;
the processing unit is used for loading the one-generation prediction model, and the one-generation prediction model is used for processing the characteristic parameters of the current dialogue data;
the output unit is used for outputting the prediction result of the single forming rate of the current dialogue data by the single forming prediction model.
7. An electronic device, comprising:
a processor and a memory;
the memory is used for storing computer operation instructions;
the processor is used for executing the method of any one of claims 1 to 6 by calling the computer operation instruction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210495977.XA CN114581162B (en) | 2022-05-09 | 2022-05-09 | Method and device for predicting order in customer service conversation process and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210495977.XA CN114581162B (en) | 2022-05-09 | 2022-05-09 | Method and device for predicting order in customer service conversation process and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114581162A CN114581162A (en) | 2022-06-03 |
| CN114581162B true CN114581162B (en) | 2022-09-02 |
Family
ID=81769127
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210495977.XA Active CN114581162B (en) | 2022-05-09 | 2022-05-09 | Method and device for predicting order in customer service conversation process and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114581162B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010239598A (en) * | 2009-03-31 | 2010-10-21 | Fujitsu Ltd | Terminal unit, and operator distribution program |
| CN104462024A (en) * | 2014-10-29 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Method and device for generating dialogue action strategy model |
| KR101644295B1 (en) * | 2015-04-16 | 2016-08-03 | (주)스윗트래커 | Integrated online customer service system and method |
| CN112102116A (en) * | 2020-09-18 | 2020-12-18 | 携程计算机技术（上海）有限公司 | Input prediction method, system, equipment and storage medium based on tourism session |
| WO2022041403A1 (en) * | 2020-08-26 | 2022-03-03 | 中山世达模型制造有限公司 | Sales order prediction method |
| CN114240495A (en) * | 2021-12-16 | 2022-03-25 | 成都新潮传媒集团有限公司 | Method and device for predicting business opportunity conversion probability and computer readable storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9661067B2 (en) * | 2013-12-23 | 2017-05-23 | 24/7 Customer, Inc. | Systems and methods for facilitating dialogue mining |
| US20150220946A1 (en) * | 2014-01-31 | 2015-08-06 | Verint Systems Ltd. | System and Method of Trend Identification |
| US20180012230A1 (en) * | 2016-07-11 | 2018-01-11 | International Business Machines Corporation | Emotion detection over social media |
| US11144980B2 (en) * | 2016-11-22 | 2021-10-12 | OrderGroove, Inc. | Adaptive scheduling of electronic messaging based on predictive consumption of the sampling of items via a networked computing platform |
| US20180189273A1 (en) * | 2016-12-23 | 2018-07-05 | OneMarket Network LLC | Maintaining context in transaction conversations |
| US11159679B2 (en) * | 2019-02-26 | 2021-10-26 | Cigna Taiwan Life Assurance Co. Ltd. | Automated systems and methods for natural language processing with speaker intention inference |
-
2022
- 2022-05-09 CN CN202210495977.XA patent/CN114581162B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010239598A (en) * | 2009-03-31 | 2010-10-21 | Fujitsu Ltd | Terminal unit, and operator distribution program |
| CN104462024A (en) * | 2014-10-29 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Method and device for generating dialogue action strategy model |
| KR101644295B1 (en) * | 2015-04-16 | 2016-08-03 | (주)스윗트래커 | Integrated online customer service system and method |
| WO2022041403A1 (en) * | 2020-08-26 | 2022-03-03 | 中山世达模型制造有限公司 | Sales order prediction method |
| CN112102116A (en) * | 2020-09-18 | 2020-12-18 | 携程计算机技术（上海）有限公司 | Input prediction method, system, equipment and storage medium based on tourism session |
| CN114240495A (en) * | 2021-12-16 | 2022-03-25 | 成都新潮传媒集团有限公司 | Method and device for predicting business opportunity conversion probability and computer readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 面向服务架构中的电子商务对话;高明;《中国金融电脑》;20070815(第08期);47-51 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114581162A (en) | 2022-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2018040944A1 (en) | System, method, and device for identifying malicious address/malicious purchase order | |
| CN110717325B (en) | Text emotion analysis method and device, electronic equipment and storage medium | |
| CN111680165B (en) | Information matching method and device, readable storage medium and electronic equipment | |
| CN110597965B (en) | Emotion polarity analysis method and device for article, electronic equipment and storage medium | |
| CN114863437B (en) | Text recognition method and device, electronic equipment and storage medium | |
| CN112712079A (en) | Character recognition method and device based on progressive coding and electronic equipment | |
| CN115935185A (en) | Training method and device for recommendation model | |
| CN111738807A (en) | Method, computing device, and computer storage medium for recommending target objects | |
| CN111444399A (en) | Reply content generation method, device, equipment and readable storage medium | |
| CN114218505A (en) | Abnormal space-time point identification method and device, electronic equipment and storage medium | |
| CN114360027A (en) | Training method and device for feature extraction network and electronic equipment | |
| CN113722588A (en) | Resource recommendation method and device and electronic equipment | |
| CN114581162B (en) | Method and device for predicting order in customer service conversation process and electronic equipment | |
| CN116739653A (en) | Sales data acquisition and analysis system and method thereof | |
| CN117056728A (en) | Time sequence generation method, device, equipment and storage medium | |
| CN112836036B (en) | Interactive training method and device for intelligent agent, terminal and storage medium | |
| CN113112326A (en) | User identification method, method for displaying data to user and related device | |
| CN114969517A (en) | Training method and recommendation method and device of object recommendation model and electronic equipment | |
| CN114420168A (en) | Emotion recognition method, device, equipment and storage medium | |
| CN114141236A (en) | Language model updating method and device, electronic equipment and storage medium | |
| CN110119770B (en) | Decision tree model construction method, device, electronic equipment and medium | |
| CN111310460B (en) | Statement adjusting method and device | |
| CN110992079A (en) | Commodity click rate prediction method based on time series filling | |
| CN112825174A (en) | Litigation prediction method, device, system and computer storage medium | |
| CN110807118A (en) | Image comment generation method and device and electronic equipment |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |