A model training method, a business execution method and device

Abstract

The specification discloses a model training method, a business execution method and device. A server acquires historical conversation records of a user and a customer service, and deletes key sentences in the historical conversation records after acquiring the historical conversation records, wherein the key sentences include sentences in which the customer service asks the user about a problem to be solved by the user. The historical conversation records after deleting the key sentences are input into a prediction model to be trained to output sentences in which the customer service predicts that the customer service trains the user about the problem to be solved by the user as first predicted sentences. The prediction model is trained with the optimization target of minimizing the deviation between the key sentences and the first predicted sentences.

Description

Technical Field

[0001] This specification relates to the field of artificial intelligence, and in particular to a method for model training, a method for business execution, and an apparatus. Background Technology

[0002] With the development of computer technology, many business operations can be performed by computers. For example, in customer service, computers can perform semantic recognition on the information sent by users to obtain their intentions and solve the problems raised by users based on their intentions.

[0003] The current form of semantic recognition mainly involves parsing the information sent by the user, obtaining the user's question through parsing, and then providing solutions to the user based on the obtained question.

[0004] However, currently, semantic recognition cannot accurately determine the problem that the user wants to solve, which increases the number of interactions between the user and customer service, reduces the efficiency of the user in performing customer service tasks, and lowers the user experience.

[0005] Therefore, accurately identifying the problems that users need to solve when performing customer service tasks and improving the efficiency of users in performing customer service tasks is an urgent problem to be solved. Summary of the Invention

[0006] This specification provides a method for model training, a method for business execution, and an apparatus to partially solve the aforementioned problems existing in the prior art.

[0007] The following technical solution is adopted in this specification:

[0008] This manual provides a method for model training, including:

[0009] Access the user's historical conversation records with customer service;

[0010] Delete key statements from the historical dialogue records, wherein the key statements include statements that indicate the problem that the user needs to solve;

[0011] The historical dialogue records after deleting the key statements are input into the prediction model to be trained, so as to output the statement predicted by the prediction model in which the customer service asks the user about the problem that the user needs to solve, as the first predicted statement.

[0012] The prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first predicted statement.

[0013] Optionally, before training the prediction model, the method further includes:

[0014] Based on the business scenario in which the prediction model is applied, select a guiding statement that matches the business scenario;

[0015] The prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first predicted statement, specifically including:

[0016] Based on the guiding statement, the prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first prediction statement.

[0017] Optionally, based on the guiding statement, the prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first prediction statement, specifically including:

[0018] With the goal of minimizing the deviation between the key statement and the first predicted statement, the prediction model is trained to obtain the prediction model to be optimized.

[0019] The historical dialogue records after deleting the key statement and the guiding statement are input into the prediction model to be trained, so as to output the statement predicted by the prediction model based on the guiding statement, in which the customer service asks the user about the problem that the user needs to solve, as the second prediction statement.

[0020] The prediction model to be optimized is trained with the goal of minimizing the deviation between the key statement and the second predicted statement.

[0021] Optionally, for any guiding statement, the encoding of the guiding statement is the same in the prediction model corresponding to different business scenarios.

[0022] This specification provides a method for performing business operations, including:

[0023] When the user is detected performing customer service tasks, the user's conversation history is obtained;

[0024] The dialogue record is input into a pre-trained prediction model to obtain the dialogue statements output by the prediction model. The prediction model is trained using the model training method described above.

[0025] Based on the dialogue statements, perform customer service for the user;

[0026] After confirming the completion of the customer service task, the prediction model is trained using the complete dialogue record.

[0027] Optionally, the method further includes:

[0028] If it is detected that after the dialogue statement is sent to the user, the user does not reply with the specified response statement corresponding to the dialogue statement, the dialogue record is sent to the device used by the human customer service representative, so that the device can perform the customer service service based on the operation performed by the human customer service representative.

[0029] This specification provides a model training apparatus, comprising:

[0030] The acquisition module is used to retrieve the user's historical conversation records with customer service.

[0031] The deletion module is used to delete key statements in the historical dialogue records, wherein the key statements include statements that indicate the problem that the user needs to solve;

[0032] The input module is used to input the historical dialogue record after deleting the key statement into the prediction model to be trained, so as to output the statement predicted by the prediction model in which the customer service asks the user about the problem that the user needs to solve, as the first predicted statement.

[0033] The training module is used to train the prediction model with the optimization objective of minimizing the deviation between the key statement and the first predicted statement.

[0034] This specification provides an apparatus for performing a business operation, comprising:

[0035] The acquisition module is used to acquire the user's conversation records when it detects that the user is performing customer service business;

[0036] The input module is used to input the dialogue record into a pre-trained prediction model and obtain the dialogue statement output by the prediction model. The prediction model is trained using the model training method described above.

[0037] An execution module is used to perform customer service tasks for the user based on the dialogue statements.

[0038] The training module is used to train the prediction model using complete dialogue records after the customer service service has been completed.

[0039] This specification provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for model training and business execution.

[0040] This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for model training and method for business execution.

[0041] The above-mentioned technical solutions adopted in this specification can achieve the following beneficial effects:

[0042] In the model training method provided in this specification, the server obtains the historical dialogue records between the user and customer service. After obtaining the historical dialogue records, key statements are deleted from the records. These key statements include statements from the customer service representative asking the user about the problem they need to solve. The historical dialogue records after deleting key statements are input into the prediction model to be trained, and the prediction model outputs the statement predicted by the model that the customer service representative asked the user about the problem they need to solve, which is used as the first predicted statement. The prediction model is trained with the optimization objective of minimizing the deviation between the key statements and the first predicted statement.

[0043] As can be seen from the above method, the server uses historical dialogue records as training samples, enabling the prediction model to predict key statements based on the entire historical dialogue record. By combining this information with the entire historical dialogue record, the model can accurately determine the problem the user needs to solve. Therefore, applying the prediction model trained using this method in practical applications can accurately predict the problem the user needs to solve, thereby improving the efficiency of customer service. Attached Figure Description

[0044] The accompanying drawings, which are included to provide a further understanding of this specification and form part of this specification, illustrate exemplary embodiments and are used to explain this specification, but do not constitute an undue limitation thereof. In the drawings:

[0045] Figure 1 This is a flowchart illustrating a model training method provided in this specification;

[0046] Figure 2 This is a flowchart illustrating one of the business execution methods provided in this specification;

[0047] Figure 3 A schematic diagram of a model training apparatus provided in this specification;

[0048] Figure 4 A schematic diagram of a business execution device provided in this specification;

[0049] Figure 5 This specification provides a corresponding Figure 1 and Figure 2 A schematic diagram of an electronic device. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0051] The technical solutions provided in the various embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0052] Figure 1 This is a flowchart illustrating a model training method provided in this specification, including the following steps:

[0053] S101: Obtain the user's historical conversation records with customer service.

[0054] To provide services to users more conveniently, business operations can be performed by computers or with computer assistance. For example, in customer service responses, computers can be used to perform semantic recognition on the information sent by users to obtain the problems that users need to solve, and then perform customer service operations based on the problems that users need to solve.

[0055] Currently, the main method for semantic recognition in customer service responses is to parse the user's message after it is sent, obtain the user's question, and then request the necessary information from the user. Once the necessary information is obtained, a solution is proposed. While this method can identify the user's question related to the message, because it performs semantic recognition on a single message, the identified question may not be the user's intended problem. It may also require the user to provide duplicate information, thus degrading the user experience.

[0056] To address the aforementioned issues, this specification provides a method for model training.

[0057] In this manual, the subject executing the model training method can be a server or other device set up on the business platform, or a terminal device such as a desktop computer or laptop computer. For ease of description, the following description will only use a server as an example to illustrate the model training method provided in this manual.

[0058] First, the server needs to obtain the user's historical dialogue records with customer service. These historical dialogue records can be any dialogue record obtained in any business scenario where the user's problem-solving needs to be clearly identified. The business scenario here can be the context in which the user performs customer service actions. For example, if the user needs customer service due to a late food delivery, then the business scenario for this customer service action is the food delivery scenario when the user enters the customer service interface through a food delivery order. This specification does not limit the business scenario for obtaining historical dialogue records; dialogue records from any business scenario can be used as historical dialogue records.

[0059] S102: Delete key statements from the historical dialogue record, wherein the key statements include statements that indicate the problem that the user needs to solve.

[0060] After retrieving historical conversation records, the server deletes key statements from these records. These key statements include those indicating the user's problem to be solved. Simultaneously, the server can use these key statements as positive samples to reduce the workload required for labeling samples.

[0061] In this specification, a key statement can refer to a statement sent by the user that clearly identifies the problem the user needs to solve. For example, if a user sends, "My food delivery order is late," the server can determine from this statement that the problem the user needs to solve is a food delivery delay. In this case, the statement is a key statement.

[0062] Of course, it can also refer to a statement sent by customer service asking the user about the problem they need to solve. For example, customer service might send: "Is the problem you need to solve a delivery delay issue?" Once the server receives the user's confirmation response to this statement (e.g., yes, etc.), it can determine that the user's problem is indeed a delivery delay. In this case, this statement can also be considered a key statement. It should be noted that the key statement here can be a single statement or multiple statements.

[0063] S103: Input the historical dialogue record after deleting the key statement into the prediction model to be trained, so as to output the statement predicted by the prediction model in which the customer service asks the user about the problem that the user needs to solve, as the first predicted statement.

[0064] The server can input the historical dialogue records after deleting key statements into the prediction model to be trained. Based on the contextual information of the historical dialogue records after deleting key statements, the prediction model can predict the problem that the user needs to solve, which was represented by the deleted key statements. The prediction model then outputs the statement representing the predicted problem that the user needs to solve, as the first predicted statement.

[0065] In this specification, when the server inputs the historical dialogue records after deleting key statements into the prediction model to be trained, pre-set guiding statements can also be input into the prediction model to be trained, so that the model to be trained can output statements representing the predicted problem that the user needs to solve based on the historical dialogue records after deleting key statements and the guiding statements.

[0066] The guiding statements can be sentiment prediction statements. By setting sentiment prediction statements, the prediction model can determine the user's emotions based on the information sent by the user, and propose appropriate dialogue methods and solutions based on the user's emotions. Because users may be emotionally agitated when raising issues with customer service, directly offering solutions in such situations often exacerbates their emotions. To improve the user experience, in cases where users are emotionally agitated, it is advisable to first calm the user and then offer solutions.

[0067] Of course, guiding statements can also be scenario-based guiding statements used to determine the context in which the user needs to resolve a problem, or event-based guiding statements used to determine the event stage to which the user needs to resolve a problem (for example, in a food delivery scenario, if a problem occurs when a user places an order, it is a problem that occurred before the food delivery process was executed; if a problem occurs during delivery, it is a problem that occurred during the food delivery process). There are many types of guiding statements, which will not be listed here.

[0068] During model training, the server can select appropriate guiding statements to add to the prediction model based on the business scenario in which the model is applied, so that the trained prediction model can be applied to that business scenario. For example, the server can add sentiment prediction statements during model training, allowing the prediction model to combine the user's emotions and the statements sent by the user to propose an appropriate dialogue approach.

[0069] In practical applications, servers need to train corresponding prediction models for each business scenario based on the data acquired in that scenario. To reduce the cost of training these prediction models, this specification stipulates that for any guiding statement, its encoding is identical across prediction models for different business scenarios. Specifically, the prediction model can include an encoding layer for encoding the guiding statement. Therefore, after training any prediction model, the trained encoding layer can be directly reused in prediction models for other business scenarios. This ensures that for guiding statements shared across all business scenarios, inputting the guiding statement into the prediction model for any business scenario will yield the same encoding. Thus, from the overall perspective, this significantly reduces the cost of model training and improves the efficiency of model training and deployment.

[0070] S104: The prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first predicted statement.

[0071] After inputting the historical dialogue records after deleting key statements into the prediction model to be trained, the prediction model is trained with the optimization objective of minimizing the deviation between the key statements and the first predicted statement.

[0072] Of course, considering the inclusion of the aforementioned guiding statements in the training process of the prediction model, there are two specific scenarios for model training. In the first scenario, the server can train a prediction model with the optimization objective of minimizing the deviation between the key statement and the first predicted statement. This trained model is not the final model, but rather a pre-trained model (hereafter referred to as the pre-trained model for ease of description). After obtaining the pre-trained model, the server can select appropriate guiding statements based on the business scenario in which the prediction model is applied. The selected guiding statements, along with the historical dialogue records after deleting the key statements, are input into the pre-trained model. The output statement, based on the guiding statements, predicts the problem the user needs to solve, serving as the second predicted statement. Then, the pre-trained model is retrained with the optimization objective of minimizing the deviation between the key statement and the second predicted statement.

[0073] In the second scenario, the server can directly input the guiding statement for the corresponding business scenario of the prediction model, along with the historical dialogue records after deleting the key statement, into the prediction model to be trained. Based on the input guiding statement, the prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first predicted statement. In this way, the trained model is the final model.

[0074] To improve model training efficiency, this manual describes several model parameter optimization techniques. For example, an adapter can be used to add extra parameters to the pre-trained prediction model. When further training the model for specific business scenarios, the original model parameters can be kept constant while the added extra parameters are adjusted, thus improving training efficiency. Other methods, such as prefix tuning, can also be used, but these will not be detailed here.

[0075] The above describes a model training method provided in this specification. The following, with reference to the accompanying drawings, illustrates how a model trained using this method performs business operations in practical applications. Figure 2 As shown.

[0076] Figure 2 The flowchart illustrating a business execution method provided in this specification includes the following steps:

[0077] S201: When the user is detected performing customer service, the user's conversation records are obtained.

[0078] When the server detects that a user needs customer service assistance, it can input the message and the previous conversation between the user and customer service representative as a dialogue record into the prediction model each time the user sends a message.

[0079] The server can also input the statement containing preset keywords (such as "return" or "refund") and the preceding dialogue records into the prediction model after capturing the statement sent by the user. Alternatively, the server can input the dialogue records into the prediction model after determining that the number of statements in the dialogue records between the user and customer service exceeds a set limit.

[0080] It should be noted that, in this manual, the customer service representative who speaks with the user during the customer service process may refer to a virtual customer service representative.

[0081] S202: Input the dialogue record into a pre-trained prediction model to obtain the dialogue statements output by the prediction model. The prediction model is trained using the model training method described above.

[0082] After the server obtains the user's dialogue records, it inputs the dialogue records into a pre-trained prediction model, obtains the dialogue statements output by the prediction model, and sends the output dialogue statements to the user.

[0083] S203: Based on the dialogue statement, perform customer service for the user.

[0084] After the server sends the dialogue statement to the user, if it confirms from the user's reply that the problem expressed in the dialogue statement is different from the problem the user needs to solve, the server can send the dialogue record to the device used by the human customer service representative, so that the human customer service representative can perform customer service tasks based on the dialogue record.

[0085] If the user's response confirms that the problem expressed in the dialogue is the same as the problem the user needs to solve, the server can propose a solution based on the dialogue. For example, the server can redirect to a page to implement the solution.

[0086] S204: After confirming that the customer service service has been completed, train the prediction model using the complete dialogue record.

[0087] After the server has completed the customer service task, it can train a prediction model using the complete conversation log.

[0088] It should be noted that, in this specification, the predictive model may also send the predicted problem to the device used by human customer service after determining that the predicted problem is the same as the problem that the user needs to solve. The device will then perform customer service based on the actions taken by the human customer service representative.

[0089] Currently, the main method for semantic recognition in customer service responses is to parse the user's message after sending it to customer service to obtain the problem the user needs to solve. Because this problem is predicted based on a single sentence sent by the user, the accuracy of the prediction is low. Furthermore, if the predicted problem is inaccurate, it cannot be corrected by considering the context of the surrounding sentences, causing the error to propagate and affecting the execution of subsequent customer service tasks. Moreover, since this method predicts the user's problem based on a single sentence, it is difficult to handle situations where the user needs to solve multiple problems simultaneously.

[0090] As can be seen from the model training method and business execution method provided in this manual, the predictive model can predict the problem that the user needs to solve based on the entire dialogue record. Thus, the predictive model trained using the above method can accurately predict the problem that the user needs to solve in practical applications, improving the efficiency of customer service operations.

[0091] In other words, the predictive model provided in this manual can predict the problems a user needs to solve based on the entire dialogue record. Thus, every sentence in the dialogue record can influence the final predicted problem, thereby achieving a globally optimal result. Furthermore, if a user needs to solve multiple problems simultaneously, the entire dialogue record can be used to predict these multiple problems, allowing customer service to be executed accordingly.

[0092] The above describes a model training method provided by one or more embodiments of this specification. Based on the same idea, this specification also provides a corresponding model training apparatus, such as... Figure 3 As shown.

[0093] Figure 3 A schematic diagram of a model training apparatus provided in this specification includes:

[0094] Module 301 is used to retrieve the user's historical conversation records with customer service.

[0095] The deletion module 302 is used to delete key statements in the historical dialogue record, wherein the key statements include statements that indicate the problem that the user needs to solve;

[0096] The input module 303 is used to input the historical dialogue record after deleting the key statement into the prediction model to be trained, so as to output the statement predicted by the prediction model in which the customer service asks the user about the problem that the user needs to solve, as the first predicted statement.

[0097] The training module 304 is used to train the prediction model with the optimization objective of minimizing the deviation between the key statement and the first prediction statement.

[0098] Optionally, before training the prediction model, the training module 304 is further configured to select a guiding statement that matches the business scenario in which the prediction model is applied.

[0099] The training module 304 is specifically used to train the prediction model based on the guiding statement, with the optimization objective of minimizing the deviation between the key statement and the first prediction statement.

[0100] Optionally, the training module 304 is specifically used to: train the prediction model with the optimization objective of minimizing the deviation between the key statement and the first predicted statement to obtain a prediction model to be optimized; input the historical dialogue record after deleting the key statement and the guiding statement into the prediction model to be trained, so as to output the statement predicted by the prediction model based on the guiding statement, in which the customer service representative asks the user about the problem that the user needs to solve, as the second predicted statement; and train the prediction model to be optimized with the optimization objective of minimizing the deviation between the key statement and the second predicted statement.

[0101] Optionally, for any guiding statement, the encoding of the guiding statement is the same in the prediction model corresponding to different business scenarios.

[0102] Figure 4 A schematic diagram of a business execution apparatus provided in this specification includes:

[0103] The acquisition module 401 is used to acquire the user's conversation records when it detects that the user is performing customer service business;

[0104] The input module 402 is used to input the dialogue record into a pre-trained prediction model and obtain the dialogue statement output by the prediction model. The prediction model is trained by the above-mentioned model training method.

[0105] Execution module 403 is used to perform customer service for the user based on the dialogue statement;

[0106] The training module 404 is used to train the prediction model using the complete dialogue record after the customer service service has been completed.

[0107] Optionally, the execution module 403 is further configured to, if it is detected that after the dialogue statement is sent to the user and the user does not reply with the specified response statement corresponding to the dialogue statement, send the dialogue record to the device used by the human customer service representative, so that the device performs the customer service service based on the operation performed by the human customer service representative.

[0108] This specification also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 and Figure 2 This provides a method for model training and a method for business execution.

[0109] This instruction manual also provides Figure 5 One of the corresponding Figure 1 and Figure 2 A schematic diagram of the structure of an electronic device. (e.g.) Figure 5At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for the business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to achieve the above-mentioned functions. Figure 1 The model training method and Figure 2 The method for executing the business described herein. Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0110] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0111] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0112] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0113] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0114] Those skilled in the art will understand that embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0115] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0118] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0119] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0120] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0121] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0122] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0124] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0125] The above description is merely an embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A method for training a model, characterized in that, include: Obtain historical dialogue records between the user and customer service; delete key statements from the historical dialogue records, wherein the key statements include statements indicating the problem the user needs to solve; select guiding statements that match the business scenario to which the prediction model is applied, the guiding statements being used to guide the prediction model to generate dialogue statements that conform to the business scenario; input the historical dialogue records after deleting the key statements and the guiding statements into the prediction model to be trained, so as to output the statement predicted by the prediction model based on the guiding statements, in which the customer service representative asks the user about the problem the user needs to solve, as the first predicted statement; The prediction model is trained with the optimization objective of minimizing the deviation between the key statement and the first predicted statement. Specifically, this includes: training the prediction model to obtain a prediction model to be optimized by minimizing the deviation between the key statement and the first predicted statement; inputting the historical dialogue record after deleting the key statement and the guiding statement into the prediction model to be optimized, and outputting the statement predicted by the prediction model based on the guiding statement, in which the customer service representative asks the user about the problem the user needs to solve, as the second predicted statement; and training the prediction model to be optimized with the optimization objective of minimizing the deviation between the key statement and the second predicted statement.

2. The method as described in claim 1, characterized in that, For any guiding statement, the encoding of the guiding statement is the same in the prediction model corresponding to different business scenarios.

3. A method for executing business operations, characterized in that, include: When a user is detected performing customer service tasks, the user's conversation history is retrieved; The dialogue record is input into a pre-trained prediction model to obtain the dialogue statement output by the prediction model, wherein the prediction model is trained by the method described in any one of claims 1 to 2; based on the dialogue statement, customer service services for the user are performed; after determining that the customer service services have been performed, the prediction model is trained using the complete dialogue record.

4. The method as described in claim 3, characterized in that, The method further includes: if it is detected that after the dialogue statement is sent to the user, the user does not reply with the specified response statement corresponding to the dialogue statement, the dialogue record is sent to the device used by the human customer service representative, so that the device performs the customer service service based on the operation performed by the human customer service representative.

5. A device for model training, characterized in that, include: The acquisition module is used to retrieve the user's historical conversation records with customer service. The system includes a deletion module for deleting key statements from the historical dialogue records, wherein the key statements include statements representing the problem that the user needs to solve; a selection module for selecting a guiding statement that matches the business scenario to which the prediction model is applied, wherein the guiding statement guides the prediction model to generate dialogue statements that conform to the business scenario; and an input module for inputting the historical dialogue records after deleting the key statements and the guiding statements into the prediction model to be trained, so as to output the statement predicted by the prediction model based on the guiding statements, in which the customer service representative asks the user about the problem that the user needs to solve, as the first predicted statement. The training module is used to train the prediction model with the optimization objective of minimizing the deviation between the key statement and the first predicted statement. Specifically, it includes: training the prediction model with the optimization objective of minimizing the deviation between the key statement and the first predicted statement to obtain a prediction model to be optimized; inputting the historical dialogue record after deleting the key statement and the guiding statement into the prediction model to be optimized to output the statement predicted by the prediction model to be optimized based on the guiding statement, which is the statement in which the customer service representative asks the user about the problem that the user needs to solve, as the second predicted statement; and training the prediction model to be optimized with the optimization objective of minimizing the deviation between the key statement and the second predicted statement.

6. An apparatus for executing business operations, characterized in that, include: The acquisition module is used to acquire users' conversation records when it detects that a user is performing customer service tasks; An input module is used to input the dialogue record into a pre-trained prediction model to obtain the dialogue statement output by the prediction model, wherein the prediction model is trained by the method described in any one of claims 1 to 2; an execution module is used to execute customer service services for the user based on the dialogue statement. The training module is used to train the prediction model using complete dialogue records after the customer service service has been completed.

7. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1-2 or 3-4.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method described in any one of claims 1 to 2 or 3 to 4.

Images