Customer service response method and device, storage medium and electronic equipment

Abstract

The invention provides a customer service response method and device, a storage medium and electronic equipment, and the method comprises the steps: obtaining first session data when a customer service response demand is detected, and the first session data comprises to-be-replied information inputted by a target customer; generating an initial strategy suggestion based on the first session data, and generating a first response result based on the first session data and the initial strategy suggestion; second session data is determined, a first strategy application judgment and optimization suggestion result is generated based on the second session data and the first response result, and the second session data comprises to-be-replied information; and based on the first strategy application judgment and optimization suggestion result and the first response result, determining target customer service response content, and outputting the target customer service response content. According to the embodiment of the invention, the response content of the target customer service can be flexibly determined, so that the response effect of the response content of the target customer service is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a customer service response method, device, storage medium and electronic device. Background Art

[0002] With the advancement of AI (Artificial Intelligence) technology, related intelligent customer service systems have been widely researched and applied. In particular, large language models (LLMs) have made significant progress in the field of natural language processing (NLP) in recent years, accelerating the development of intelligent customer service. However, related technologies often directly generate customer service responses through generative models, resulting in overly mechanical responses and, consequently, poor customer service response effectiveness. Therefore, there is currently no effective solution for flexibly determining customer service response content to improve its effectiveness. Summary of the Invention

[0003] In view of this, the embodiments of the present invention provide a customer service response method, device, storage medium and electronic device to solve the problems that the customer service response content generated by related technologies is too mechanical and the effect is poor; that is, the embodiments of the present invention can generate initial strategy suggestions and introduce the initial strategy suggestions into the generation process of the response results to achieve the introduction of communication and emotion regulation skills through the initial strategy suggestions, which can comprehensively improve the intelligence and humanization of customer service responses, thereby flexibly determining the target customer service response content, so as to effectively improve the response effect of the target customer service response content, and thus effectively improve the customer experience.

[0004] According to one aspect of an embodiment of the present invention, a customer service response method is provided, the method comprising:

[0005] When a customer service response requirement is detected, first session data is obtained, where the first session data includes information to be replied to entered by the target customer;

[0006] generating an initial policy suggestion based on the first session data, and generating a first response result based on the first session data and the initial policy suggestion;

[0007] Determining second session data, and generating a first strategy application evaluation and optimization recommendation result based on the second session data and the first response result, wherein the second session data includes the information to be replied;

[0008] Based on the first strategy application evaluation and optimization suggestion results and the first response result, target customer service response content is determined, and the target customer service response content is output.

[0009] According to another aspect of an embodiment of the present invention, a customer service response device is provided, the device comprising:

[0010] an acquiring unit, configured to acquire first session data when a customer service response requirement is detected, wherein the first session data includes information to be replied to input by the target customer;

[0011] a processing unit, configured to generate an initial policy suggestion based on the first session data, and generate a first response result based on the first session data and the initial policy suggestion;

[0012] The processing unit is further configured to determine second session data, and generate a first policy application evaluation and optimization recommendation result based on the second session data and the first response result, wherein the second session data includes the information to be replied;

[0013] The processing unit is further configured to determine target customer service response content based on the first strategy application evaluation and optimization suggestion result and the first response result, and output the target customer service response content.

[0014] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising a processor and a memory storing a program, wherein the program comprises instructions, which, when executed by the processor, cause the processor to perform the above-mentioned method.

[0015] According to another aspect of an embodiment of the present invention, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the above-mentioned method.

[0016] When a customer service response request is detected, embodiments of the present invention can obtain first session data, the first session data including the information to be replied to entered by the target customer; generate an initial strategy recommendation based on the first session data, and generate a first response result based on the first session data and the initial strategy recommendation. Then, second session data can be determined, and based on the second session data and the first response result, a first strategy application evaluation and optimization recommendation result can be generated, the second session data including the information to be replied to. Furthermore, the target customer service response content can be determined based on the first strategy application evaluation and optimization recommendation result and the first response result, and the target customer service response content can be output. It can be seen that embodiments of the present invention can achieve the comprehensive improvement of the intelligence and humanization of customer service responses by introducing communication and emotion regulation skills through the initial strategy recommendation by generating an initial strategy recommendation and introducing the initial strategy recommendation into the response result generation process. This allows for the flexible determination of the target customer service response content, effectively improving the response effectiveness of the target customer service response content, and thereby effectively enhancing the customer experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Further details, features and advantages of the present invention are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 A schematic flow chart of a customer service response method according to an exemplary embodiment of the present invention is shown;

[0019] Figure 2 A schematic flow chart showing another customer service response method according to an exemplary embodiment of the present invention is shown;

[0020] Figure 3 A schematic flow chart showing another customer service response method according to an exemplary embodiment of the present invention is shown;

[0021] Figure 4 A schematic block diagram of a customer service answering device according to an exemplary embodiment of the present invention is shown;

[0022] Figure 5 A block diagram of an exemplary electronic device capable of implementing the embodiments of the present invention is shown. DETAILED DESCRIPTION

[0023] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0024] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0025] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0026] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0027] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0028] It should be noted that the execution subject of the customer service response method provided in the embodiment of the present invention can be one or more electronic devices, and the embodiment of the present invention does not limit this; wherein, the electronic device can be a terminal (i.e., a client) or a server. Then, when the execution subject includes multiple electronic devices, and the multiple electronic devices include at least one terminal and at least one server, the customer service response method provided in the embodiment of the present invention can be jointly executed by the terminal and the server. Accordingly, the terminals mentioned here can include but are not limited to: smart phones, tablet computers, laptops, desktop computers, intelligent voice interaction devices, etc. The server mentioned here can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing (cloud computing), cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and basic cloud computing services such as big data and artificial intelligence platforms, etc.

[0029] Optionally, an embodiment of the present invention may also provide a customer service system, which may be deployed on one or more electronic devices. In this case, the execution entity of the customer service response method provided by the embodiment of the present invention may be the customer service system, that is, the embodiment of the present invention may execute the customer service response method through the customer service system in the electronic device, and so on.

[0030] Based on the above description, an embodiment of the present invention proposes a customer service response method, which can be executed by the electronic device (terminal or server) mentioned above; or, the customer service response method can be executed by the terminal and the server together, etc. For the sake of convenience, the following description will be based on an example of an electronic device executing the customer service response method; Figure 1 As shown, the customer service response method may include the following steps S101-S104:

[0031] S101, when a customer service response requirement is detected, first session data is obtained, where the first session data includes information to be replied to input by a target customer.

[0032] Optionally, the electronic device may determine that a customer service response request has been detected upon detecting a pending reply message entered by the target customer. Alternatively, in an intelligent customer service scenario, the electronic device may determine that a customer service response request has been detected upon detecting a pending reply message entered by the target customer. Alternatively, in a manual customer service scenario, the electronic device may determine that a customer service response request has been detected upon detecting a customer service response generation instruction for the pending reply message, and so on. This is not limited in the present embodiment. Optionally, the target customer may be any customer, and this is not limited in the present embodiment. Optionally, the pending reply message may be any message, and this is not limited in the present embodiment. Optionally, the electronic device may determine that a customer service response generation instruction has been detected upon detecting a target customer service representative performing a customer service response generation operation in response to the pending reply message. Alternatively, the electronic device may determine that a customer service response generation instruction has been detected by receiving a customer service response generation instruction sent by the front end, and so on. It should be noted that the present embodiment does not limit the specific implementation of the customer service response generation operation. Optionally, the customer service response generation instruction may carry the pending reply message. Optionally, the target customer service representative may be any customer service representative, and this is not limited in the present embodiment.

[0033] Optionally, the first session data may further include a first historical session; optionally, the first historical session may include the target customer's historical session, or may include all historical sessions where the difference between the session time and the current system time is less than a first preset time difference threshold, etc.; optionally, the first historical session may be set based on experience or actual needs, which is not limited in this embodiment of the present invention. Optionally, the first preset time difference threshold may be set based on experience or actual needs, which is not limited in this embodiment of the present invention.

[0034] In this embodiment of the present invention, the first session data may be obtained in the following ways, but is not limited to:

[0035] The first acquisition method: the first session data is stored in the storage space of the electronic device. In this case, the electronic device can acquire the first session data from the storage space of the electronic device.

[0036] The second acquisition method: The electronic device can add the information to be replied to to the first session data, and obtain the first historical session download link to download the first historical session based on the first historical session download link, thereby adding the first historical session to the first session data to obtain the first session data, and so on.

[0037] S102: Generate an initial policy suggestion based on the first session data, and generate a first response result based on the first session data and the initial policy suggestion.

[0038] Optionally, the electronic device may call a first response result generation model to generate a first response result based on the first session data and the initial policy recommendation; that is, the electronic device may determine first response result generation model input data based on the first session data and the initial policy recommendation, and input the first response result generation model input data into the first response result generation model to output the first response result through the first response result generation model. Optionally, when determining the first response result generation model input data based on the first session data and the initial policy recommendation, the electronic device may use the first session data and the initial policy recommendation as the first response result generation model input data, that is, the first response result generation model input data may be input into the first response result generation model at this time; or, the first session data, the initial policy recommendation, and the first preset result generation prompt information may be used to construct the first response result generation model input data; or, the customer service goal may be determined, and the first session data, the initial policy recommendation, the customer service goal, and the first preset result generation prompt information may be used to construct the first response result generation model input data, and so on; the embodiments of the present invention are not limited to this. Optionally, the first preset result generation prompt information may be set based on experience or actual needs, and this embodiment of the present invention is not limited thereto. Optionally, the first preset result generation prompt information may also include customer service goals, etc. Optionally, the number of strategy recommendations in the initial strategy recommendation may be one or more, and this embodiment of the present invention is not limited thereto.

[0039] Optionally, the customer service target may be input by the target customer service through a customer service response generation operation. In this case, the customer service response generation instruction may also carry the customer service target, thereby enabling the electronic device to determine the customer service target. Alternatively, the customer service target may be determined based on the information to be replied to. For example, a probability value of the information to be replied to for each preset customer service target among multiple preset customer service targets may be determined, and the preset customer service target with the largest probability value may be used as the customer service target. Alternatively, the information to be replied to may be segmented to obtain at least one segmentation result to be replied to. When a customer service target tag segmentation result exists in at least one segmentation result to be replied to, an initial customer service target corresponding to the customer service target tag segmentation result is determined from multiple initial customer service target tag pairs, and the initial customer service target corresponding to the customer service target tag segmentation result is used as the customer service target. The customer service target tag segmentation result may refer to a segmentation result existing in multiple initial customer service target tag pairs. An initial customer service target tag pair may include a segmentation and an initial customer service target corresponding to the segmentation. For example, the initial customer service target corresponding to the segmentation "cancel order" may be "keep the order as much as possible to maximize platform benefits," etc. Alternatively, the customer service target may be set based on experience or actual needs. For example, different customer service targets may be set for different business stages, etc. This is not limited in the embodiments of the present invention. Optionally, multiple preset customer service goals, multiple initial customer service goal tag pairs, and different customer service goals formulated for different business stages can be set based on experience or actual needs, and the embodiments of the present invention do not limit this.

[0040] Optionally, a response result generation model (such as a first response result generation model) can be any large language model, and the embodiment of the present invention does not limit this; that is, the embodiment of the present invention does not limit the specific model structure of the response result generation model such as the first response result generation model; optionally, a response result generation model can be set according to experience or actual needs, or it can be obtained by model training through a response training data set, and the embodiment of the present invention does not limit this. Exemplarily, the electronic device can use the response training data set corresponding to the first response result generation model to perform model training on the first initial response result generation model to obtain a first response result generation model; optionally, a response training data can include but is not limited to a first training session data, a training strategy recommendation, and a response result label, etc., and the embodiment of the present invention does not limit this; that is, the response training data set corresponding to a response result generation model can be set according to experience or actual needs, and the embodiment of the present invention does not limit this.

[0041] Based on this, the embodiment of the present invention can incorporate initial strategy suggestions into the process of generating the first response result, so that the first response result generation model can generate a more humane response result, such as Figure 2 As shown by arrow 2 in .

[0042] S103: Determine second session data, and generate a first strategy application evaluation and optimization suggestion result based on the second session data and the first response result, wherein the second session data includes information to be replied.

[0043] Optionally, the second session data may further include a second historical session; optionally, the second historical session may include the target customer's historical session, or may include all historical sessions where the difference between the session time and the current system time is less than a second preset time difference threshold, and so on; optionally, the second historical session may be set based on experience or actual needs, which is not limited in this embodiment of the present invention. Optionally, the second preset time difference threshold may be set based on experience or actual needs, which is not limited in this embodiment of the present invention. In this embodiment of the present invention, the second session data may be the same as or different from the first session data, which is not limited in this embodiment of the present invention.

[0044] Optionally, a policy application evaluation and optimization suggestion result (such as the first policy application evaluation and optimization suggestion result) may include but is not limited to at least one of the following: a policy application score of each policy suggestion in at least one policy suggestion (such as the policy application score of a policy suggestion in the initial policy suggestion for the first response result, which may also be referred to as a policy score) and an optimization suggestion (which may also be referred to as a policy application optimization suggestion), etc., which is not limited in the embodiment of the present invention; illustratively, for the first policy application evaluation and optimization suggestion result, the at least one policy suggestion may include all policy suggestions in the initial policy suggestions, or may include any policy suggestion in the initial policy suggestions, or may include the first policy suggestion in the initial policy suggestions (such as the policy suggestion with the largest policy suggestion probability value), etc. Optionally, the policy application score of a policy suggestion may be any policy application score in the policy application score set, which is not limited in the embodiment of the present invention. Optionally, the policy application score set can be set according to experience or according to actual needs, and the embodiments of the present invention are not limited to this. For example, it is assumed that the policy application score set may include a policy application score of 1-3 points, where 1 point may indicate that the policy recommendation is poorly applied or the relevant policy recommendation is not used, 2 points may represent that the policy recommendation is basically applied but there are still obvious shortcomings, 3 points may indicate that the application is in line with the policy recommendation and can be used as a response output, and so on.

[0045] Optionally, when generating the first policy application evaluation and optimization suggestion result based on the second session data and the first response result, the electronic device may generate the first policy application evaluation and optimization suggestion result based on the second session data, the first response result and the initial policy suggestion, and so on. Since the second session data includes information to be replied, the embodiment of the present invention may generate the first policy application evaluation and optimization suggestion result based on the information to be replied (such as Figure 2 Information entered by the customer in the first response result (such as Figure 2 The response result pointed by arrow 2 in the figure) and the initial strategy suggestion (such as Figure 2 The strategy suggestion pointed by arrow 1 in the middle) generates the first strategy application evaluation and optimization suggestion results, such as Figure 2 As shown by arrows 3, 4, and 5 in FIG. Based on this, the embodiment of the present invention can perform strategy application evaluation on the generated first response result and provide optimization suggestions to facilitate subsequent optimization of the response result or output of a better response result, etc.

[0046] S104: Based on the first strategy application evaluation and optimization suggestion result and the first response result, determine the target customer service response content, and output the target customer service response content.

[0047] Optionally, when outputting the target customer service response content, the electronic device may send the target customer service response content to the target customer, such as sending it to the front-end device used by the target customer, so that the front-end device used by the target customer displays the target customer service response content, and the target customer service response content can be displayed in the target customer's customer service interface; exemplarily, in the intelligent customer service scenario, the target customer service response content can be directly sent to the target customer.

[0048] Alternatively, in a manual customer service scenario, the target customer service response content can be sent to the target customer service. The target customer service can be any customer service who provides manual customer service to the target customer. For example, the target customer service response content can be displayed on the front-end device used by the target customer service. At this time, the target customer service response content can be used to provide a response reference to the target customer service, etc.; the embodiment of the present invention is not limited to this.

[0049] Optionally, in other embodiments, the first response result may also be the response result input by the target customer service in response to the information to be replied and the initial strategy suggestion, that is, the first response result may be the response result created by the manual customer service, that is, after the electronic device obtains the initial strategy suggestion, it may also send the initial strategy suggestion to the target customer service, and then the target customer service may create text according to the initial strategy suggestion to input the first response result; further, after detecting the first response result, the electronic device may generate a first strategy application evaluation and optimization suggestion result based on the second session data and the first response result, and send the first strategy application evaluation and optimization suggestion result to the target customer service, so as to provide the target customer service with a scoring basis and guidance optimization suggestions through the first strategy application evaluation and optimization suggestion result, etc.; the present invention is not limited to this. Based on this, the present invention is of great help to the improvement of manual customer service skills. It not only has theoretical strategy suggestions, but also combines specific cases and the manual customer service's own creative content for optimization creation, which reduces the entry threshold of manual customer service in this field and can improve the business capabilities of manual customer service.

[0050] In summary, it can be seen that the customer service response method proposed by the present invention can comprehensively improve the intelligence, humanity and efficiency of the customer service system. It does not blindly pursue the response speed or accuracy of task processing, but pays more attention to the use of communication skills and strategies to improve customer experience. It can also directly assist manual customer service and make reasonable prompts and improvements to the communication methods of manual customer service, which can assist manual customer service in improving skills and improving the communication skills of manual customer service in complex scenarios, so there is no excessive requirement for the prior knowledge of manual customer service.

[0051] When a customer service response request is detected, embodiments of the present invention can obtain first session data, the first session data including the information to be replied to entered by the target customer; generate an initial strategy recommendation based on the first session data, and generate a first response result based on the first session data and the initial strategy recommendation. Then, second session data can be determined, and based on the second session data and the first response result, a first strategy application evaluation and optimization recommendation result can be generated, the second session data including the information to be replied to. Furthermore, the target customer service response content can be determined based on the first strategy application evaluation and optimization recommendation result and the first response result, and the target customer service response content can be output. It can be seen that embodiments of the present invention can achieve the comprehensive improvement of the intelligence and humanization of customer service responses by introducing communication and emotion regulation skills through the initial strategy recommendation by generating an initial strategy recommendation and introducing the initial strategy recommendation into the response result generation process. This allows for the flexible determination of the target customer service response content, effectively improving the response effectiveness of the target customer service response content, and thereby effectively enhancing the customer experience.

[0052] Based on the above description, another customer service response method is proposed in the embodiment of the present invention. Accordingly, the customer service response method can be executed by the electronic device (terminal or server) mentioned above; or, the customer service response method can be executed by the terminal and the server together, and so on. For the sake of convenience, the following description will take the electronic device executing the customer service response method as an example; please refer to Figure 3 The customer service response method may include the following steps S301-S305:

[0053] S301, when a customer service response requirement is detected, first session data is obtained, where the first session data includes information to be replied to input by the target customer.

[0054] S302: Generate an initial policy suggestion based on the first session data, and generate a first response result based on the first session data and the initial policy suggestion.

[0055] In one embodiment, the initial policy recommendation can be generated based on a policy recommendation library, which includes multiple policy recommendations, and each policy recommendation includes a policy recommendation identifier and an explanation of the corresponding policy recommendation. Based on this, when generating the initial policy recommendation based on the first session data, the electronic device can call the first policy recommendation generation model, and based on the first session data, generate a policy recommendation probability value for each policy recommendation under the first session data; accordingly, based on the policy recommendation probability value of each policy recommendation under the first session data, the top M policy recommendations with the largest policy recommendation probability values can be determined from the policy recommendation library, and the M policy recommendations can be used as the initial policy recommendations to achieve the generation of the initial policy recommendation based on the first session data, where M is a positive integer. Optionally, a policy recommendation generation model (such as the first policy recommendation generation model) can be any large language model, and the embodiment of the present invention does not limit this; that is, the embodiment of the present invention does not limit the specific model structure of a policy recommendation generation model; optionally, a policy recommendation generation model can be set according to experience or actual needs, or it can be obtained by model training through a policy recommendation training data set. For example, the electronic device may use a first policy recommendation training data set to perform model training on a first initial policy recommendation generation model to obtain a first policy recommendation generation model. Optionally, a first policy recommendation training data set may include, but is not limited to, a first training session data set and a policy recommendation label, etc., which is not limited in this embodiment of the present invention. Optionally, a policy recommendation training data set may be set based on experience or actual needs, which is not limited in this embodiment of the present invention.

[0056] Optionally, when invoking the first policy suggestion generation model to generate policy suggestion probability values for each policy suggestion under the first session data based on the first session data, the electronic device may determine first policy suggestion generation input data based on the first session data and input the first policy suggestion generation input data into the first policy suggestion generation model so that the first policy suggestion generation model can output policy suggestion probability values for each policy suggestion under the first session data, thereby generating policy suggestion probability values for each policy suggestion under the first session data. Optionally, when determining the first policy suggestion generation input data based on the first session data, the first session data may be used as the first policy suggestion generation input data. In this case, the first session data may be input into the first policy suggestion generation model to generate policy suggestion probability values for each policy suggestion under the first session data. Alternatively, first preset policy suggestion generation prompt information may be determined and the first session data and the first preset policy suggestion generation prompt information may be used to construct the first policy suggestion generation input data. In this case, the first policy suggestion generation input data may include the first session data and the first preset policy suggestion generation prompt information. Alternatively, the first policy suggestion generation input data may be constructed using the first session data, the customer service goal, and the first preset policy suggestion generation prompt information, and so on. This embodiment of the present invention is not limited to this. Optionally, the first preset strategy suggestion generation prompt information can be set according to experience or according to actual needs, and the embodiment of the present invention does not limit this; optionally, the first preset strategy suggestion generation prompt information can also include customer service goals, etc.

[0057] Optionally, the electronic device may also obtain third session data, which may include customer service data, which may include at least one customer service response content; and call the second strategy recommendation generation model to determine the annotation label and explanation for each customer service decomposition statement in the customer service data based on the third session data, to obtain multiple annotation labels and explanations for each of the multiple annotation labels, where the annotation label of a customer service decomposition statement is one of the multiple annotation labels, and the explanation of a customer service decomposition statement is the explanation of the annotation label of the corresponding customer service decomposition statement; thereby constructing a strategy recommendation library based on the multiple annotation labels and the explanations for each annotation label. Optionally, the third session data may include any session data accumulated by the platform, that is, the third session data may be set based on experience or actual needs, which is not limited in this embodiment of the present invention; that is, the third session data may include a third historical session, which may be set based on experience or actual needs, which is not limited in this embodiment of the present invention. Optionally, the second strategy suggestion generation model can be set according to experience or actual needs, or it can be obtained by training the second initial strategy suggestion generation model through the second strategy suggestion training data set. The embodiment of the present invention does not limit this. Exemplarily, a second strategy suggestion training data may include but is not limited to a second training session data, a training annotation label and training explanation for each customer service decomposition statement in the corresponding second training session data, etc. The embodiment of the present invention does not limit this.

[0058] Optionally, when invoking the second policy suggestion generation model and determining the label and explanation for each customer service breakdown statement in the customer service data based on the third session data, the electronic device may determine second policy suggestion generation input data based on the third session data, and input the second policy suggestion generation input data into the second policy suggestion generation model so that the second policy suggestion generation model outputs the label and explanation for each customer service breakdown statement in the customer service data. The second policy suggestion generation input data may include the third session data. Optionally, when determining the second policy suggestion generation input data based on the third session data, the electronic device may use the third session data as the second policy suggestion generation input data. Alternatively, the electronic device may determine second preset policy suggestion generation prompt information and use the third session data and the second preset policy suggestion generation prompt information to construct the second policy suggestion generation input data. In this case, the second policy suggestion generation input data may include the third session data and the second preset policy suggestion generation prompt information, and so on. This is not limited in the embodiments of the present invention. Optionally, the second preset policy suggestion generation prompt information may be set based on experience or actual needs, and this is not limited in the embodiments of the present invention. For example, assuming that the input data for generating the second strategy suggestion may include the third conversation data and the second preset strategy suggestion generation prompt information, the second strategy suggestion generation input data may be represented as "{{third conversation data}}. The second preset strategy suggestion generation prompt information." For example, the second preset strategy suggestion generation prompt information may include: This is a conversation between a human customer service representative and a customer in the xxx field. Please deconstruct each customer service representative's response data. The customer service representative data begins with xxx. Please label each customer service representative's statement from the perspective of communication skills. The labels need to be customized (defined from the perspective of communication skills) and explained, etc. Optionally, the customer service data may include one or more customer service representative response data, which is not limited in this embodiment of the present invention. Optionally, the second preset strategy suggestion generation prompt information may also include a preset tag set, so that the tag label of any generated customer service deconstruction statement is a preset tag in the preset tag set. In this case, the multiple tag labels may include all tags in the preset tag set, etc. For example, the preset tag set may include, but is not limited to, expressing apology, empathy, customer care, statement of facts, transparency, information provision, proactive proposal, etc., which is not limited in this embodiment of the present invention. Optionally, when the second preset policy suggestion generation prompt information does not include a preset tag set, the multiple annotation tags may be generated by the second policy suggestion generation model. Optionally, the preset tag set may be set based on experience or actual needs, and this is not limited in this embodiment of the present invention.

[0059] Based on this, an embodiment of the present invention can input the third session data into the second strategy suggestion generation model through the second strategy suggestion generation input data, so that the customer service data can be disassembled through the second strategy suggestion generation model (thereby obtaining each customer service disassembly statement in the customer service data) and performing fine-grained labeling, that is, each customer service disassembly statement in the customer service data (that is, every sentence of the customer service) can be labeled and explained to determine the labeling label and explanation of each customer service disassembly statement in the customer service data.

[0060] Optionally, when constructing a policy recommendation library based on multiple annotation tags and explanations of each annotation tag, the electronic device may respectively perform vector representation on each annotation tag in the multiple annotation tags to obtain the label embedding vector of each annotation tag, and determine the target embedding vector corresponding to the corresponding annotation tag based on the label embedding vector of each annotation tag; then, based on the target embedding vector corresponding to each annotation tag, label filtering processing may be performed on the multiple annotation tags to obtain at least one filtered annotation tag, and the explanations of each filtered annotation tag in at least one filtered annotation tag may be determined from the explanations of each annotation tag; based on this, at least one filtered annotation tag and the explanations of each filtered annotation tag may be used to construct a policy recommendation library (that is, at least one filtered annotation tag and the explanations of each filtered annotation tag may be added to the policy recommendation library) so that a filtered annotation tag and the explanation of the corresponding filtered annotation tag serve as a policy recommendation in the policy recommendation library.

[0061] Optionally, when multiple annotation labels are generated by the second policy suggestion generation model, the above-mentioned vector representation of each annotation label in the multiple annotation labels can be triggered to implement label screening processing for the multiple annotation labels, thereby using at least one screened annotation label and the explanation of each screened annotation label to construct a policy suggestion library; when the multiple annotation labels are a preset label set (that is, the multiple annotation labels include all labels in the preset label set, that is, the multiple annotation labels are determined by the preset label set), the multiple annotation labels and the explanation of each annotation label can be added to the policy suggestion library (that is, the multiple annotation labels and the explanation of each annotation label can be directly used to construct the policy suggestion library), so that a annotation label and the explanation of the corresponding annotation label are used as a policy suggestion in the policy suggestion library, etc. Alternatively, no matter how the multiple annotation labels are determined, the electronic device can trigger the above-mentioned vector representation of each annotation label in the multiple annotation labels to implement label screening processing for the multiple annotation labels, etc.; the embodiments of the present invention are not limited to this. It should be understood that the annotation tags generated by the large language model may contain a large number of similar tags and noise tags. The embodiments of the present invention can improve the accuracy of at least one filtered annotation tag through tag filtering processing, thereby improving the accuracy of the strategy recommendation library.

[0062] Optionally, the electronic device may use a target text embedding model to perform a vector representation on each annotation tag; optionally, the target text embedding model may be any text embedding model, which is not limited in the embodiment of the present invention; illustratively, the target text embedding model may be a Phrase-BERT text embedding model (Improved Phrase Embeddings from BERT with an Application to Corpus Exploration, a model based on BERT (Bidirectional Encoder Representation from Transformers, a bidirectional language model)), etc. Optionally, when determining the target embedding vector corresponding to the corresponding annotation tag based on the tag embedding vector of each annotation tag, for any annotation tag among the multiple annotation tags, the tag embedding vector of any annotation tag may be used as the target embedding vector corresponding to any annotation tag; alternatively, the sentence embedding vector of the customer service decomposition sentence annotated by any annotation tag may be obtained, and the tag embedding vector of any annotation tag and the sentence embedding vector of the customer service decomposition sentence annotated by any annotation tag may be concatenated to obtain the target embedding vector corresponding to any annotation tag, etc.; the embodiment of the present invention is not limited in this regard.

[0063] Optionally, when label screening is performed on multiple label tags based on the target embedding vectors corresponding to each label tag to obtain at least one screened label tag, the electronic device may cluster the multiple label tags based on the target embedding vectors corresponding to each label tag through a target clustering algorithm to obtain at least one cluster cluster, and determine a cluster label tag from each cluster in the at least one cluster cluster, thereby obtaining at least one cluster label tag, so as to determine at least one screened label tag based on the at least one cluster label tag, wherein a cluster label tag is a label tag in a cluster cluster that is closest to the cluster center of the corresponding cluster cluster. Optionally, the target clustering algorithm may be any clustering algorithm, which is not limited in the embodiment of the present invention; illustratively, the target clustering algorithm may be a DBSCAN (Density-Based Spatial Clustering of Application with Noise) algorithm (a density-based clustering method), such as clustering the label tags using the DBSCAN algorithm with a given semantic similarity threshold (such as 0.8, etc.), etc. Optionally, when determining at least one screening label based on at least one cluster label, the at least one cluster label can be used as the at least one screening label, or the top Q cluster label containing the largest number of labels can be selected from the at least one cluster label, and the top Q cluster label used as the at least one screening label, where Q is a positive integer, etc. This is not limited in this embodiment of the present invention. Based on this, the embodiment of the present invention can implement label filtering, and label filtering can be based on frequency.

[0064] Optionally, the embodiment of the present invention can also manually verify and adjust the constructed strategy suggestion library to achieve the update of the strategy suggestion library; based on this, the embodiment of the present invention can combine the knowledge and experience of manual customer service to maintain a high-quality strategy suggestion library. Based on this, the embodiment of the present invention can introduce communication strategy suggestions to make the entire conversation communication smoother and strive to maximize the benefits of the platform. For example, a high-quality strategy knowledge base (i.e., strategy suggestion library) can be constructed by combining a large language model with manual verification and adjustment to generate high-quality initial strategy suggestions. In an embodiment of the present invention, in order to make the generation of response results more flexible and diversified, a specific solution design or specific wording expression may not be given in the strategy suggestion, but only a general strategy suggestion description may be provided, thereby improving the flexibility and diversity of the generation of response results.

[0065] In another embodiment, the electronic device may also call a third policy suggestion generation model to generate an initial policy suggestion based on the first session data without building a policy suggestion library. Optionally, the third policy suggestion generation model may be set according to experience or actual needs, or may be obtained by training a third initial policy suggestion generation model through a third policy suggestion training data set, which is not limited in this embodiment of the present invention; optionally, a third policy suggestion training data may include but is not limited to a third training session data (which may be the same as or different from the first training session data, which is not limited in this embodiment of the present invention) and a policy suggestion label (i.e., a labeled policy suggestion), etc., which is not limited in this embodiment of the present invention.

[0066] Optionally, when calling the third policy suggestion generation model to generate an initial policy suggestion based on the first session data, the electronic device may determine the third policy suggestion generation input data based on the first session data, and input the third policy suggestion generation input data into the third policy suggestion generation model to output the initial policy suggestion through the third policy suggestion generation model. Optionally, when determining the third policy suggestion generation input data based on the first session data, the first session data may be used as the third policy suggestion generation input data, in which case the first session data may be input into the third policy suggestion generation model to generate the initial policy suggestion; alternatively, a third preset policy suggestion generation prompt information may be determined, and the first session data and the third preset policy suggestion generation prompt information may be used to construct the third policy suggestion generation input data, in which case the third policy suggestion generation input data may include the first session data and the third preset policy suggestion generation prompt information; alternatively, the first session data, the customer service goal, and the third preset policy suggestion generation prompt information may be used to construct the third policy suggestion generation input data, and so on; the embodiments of the present invention are not limited to this. Optionally, the third preset strategy suggestion generation prompt information can be set according to experience or according to actual needs, and the embodiment of the present invention does not limit this; optionally, the third preset strategy suggestion generation prompt information can also include customer service goals, etc.

[0067] S303: Determine the second conversation data and determine at least one evaluation comparison example pair, where one evaluation comparison example pair includes an evaluation response content and evaluation suggestion information corresponding to the evaluation response content.

[0068] It should be noted that the embodiment of the present invention does not limit the method for determining the second session data. For example, the second session data can be obtained from a local storage space or the second historical session can be downloaded based on a second historical session download link to determine the second session data, etc.; the embodiment of the present invention does not limit this.

[0069] Optionally, the electronic device may also obtain multiple evaluation response contents corresponding to each of the multiple strategy suggestions; then, the evaluation suggestion information of each evaluation response content corresponding to each strategy suggestion may be determined respectively, and based on each evaluation response content corresponding to each strategy suggestion and the evaluation suggestion information of each evaluation response content corresponding to each strategy suggestion, multiple evaluation comparison example pairs corresponding to each strategy suggestion may be constructed. A evaluation comparison example pair corresponding to each strategy suggestion includes a evaluation response content corresponding to the corresponding strategy suggestion and the evaluation suggestion information of the corresponding evaluation response content. Optionally, the multiple evaluation response contents corresponding to a strategy suggestion may be set based on experience or actual needs, and the embodiment of the present invention does not limit this. Optionally, a evaluation response content may be a customer service response content. It should be noted that the embodiment of the present invention does not limit the method for obtaining the multiple evaluation response contents corresponding to each strategy suggestion. For example, it may be obtained from its own storage space, or it may be obtained based on a download link for the evaluation response content, etc. Among them, each evaluation response content corresponding to a strategy suggestion is each evaluation response content among the multiple evaluation response contents corresponding to a strategy suggestion.

[0070] Optionally, the evaluation suggestion information of an evaluation response content may include but is not limited to at least one of the following: a strategy application score and optimization suggestion for the corresponding evaluation response content, etc.; this is not limited in the embodiment of the present invention. Optionally, the evaluation suggestion information of an evaluation response content may be set according to experience or actual needs (such as can be determined by manual annotation), or it may be generated by a evaluation suggestion information generation model, which is not limited in the embodiment of the present invention. Optionally, the evaluation suggestion information generation model can be any large language model, which is not limited in the embodiment of the present invention; illustratively, any evaluation response content can be input into the evaluation suggestion information generation model to output the evaluation suggestion information of any evaluation response content through the evaluation suggestion information generation model, so as to determine the evaluation suggestion information of any evaluation response content, etc.

[0071] Based on this, when determining at least one evaluation and comparison example pair, the electronic device may determine at least one evaluation and comparison example pair from a plurality of evaluation and comparison example pairs corresponding to each policy suggestion. In a specific implementation, multiple evaluation and comparison example pairs corresponding to the initial policy suggestion can be determined from the multiple evaluation and comparison example pairs corresponding to each policy suggestion, as at least one evaluation and comparison example pair, that is, at this time, the at least one evaluation and comparison example pair may include multiple evaluation and comparison example pairs corresponding to the initial policy suggestion (that is, multiple evaluation and comparison example pairs corresponding to each policy suggestion in the initial policy suggestion); in another specific implementation, the multiple evaluation and comparison example pairs corresponding to each policy suggestion can be used as at least one evaluation and comparison example pair, that is, at this time, the at least one evaluation and comparison example pair may include multiple evaluation and comparison example pairs corresponding to each policy suggestion; in another specific implementation, multiple evaluation and comparison example pairs corresponding to the policy suggestion with the largest policy suggestion probability value under the first session data can be determined from the multiple evaluation and comparison example pairs corresponding to each policy suggestion, so that the determined evaluation and comparison example pairs can be used as at least one multiple evaluation and comparison example pair, and at this time, at least one multiple evaluation and comparison example pair may include multiple evaluation and comparison examples corresponding to the policy suggestion with the largest policy suggestion probability value under the first session data, and so on; the embodiments of the present invention are not limited to this.

[0072] Optionally, at least one evaluation comparison example pair may also be set according to experience or actual needs, etc.; this is not limited in the embodiment of the present invention.

[0073] In an embodiment of the present invention, the evaluation comparison example pairs can realize the effective example display of the strategy suggestion evaluation, so that the evaluation of the response result can be made more effective through the evaluation comparison example pairs, which can improve the accuracy of the strategy application evaluation and optimization suggestion results, and can assist the strategy application evaluation and optimization suggestion generation model to independently distinguish the differences between the scores (i.e., the strategy application score) to improve the model performance of the strategy application evaluation and optimization suggestion generation model.

[0074] S304: Call the first strategy application evaluation and optimization suggestion generation model to generate a first strategy application evaluation and optimization suggestion result based on the second session data, the first response result, and at least one evaluation and comparison example pair.

[0075] Optionally, a strategy application judgment and optimization suggestion generation model (such as the first strategy application judgment and optimization suggestion generation model) can be any large language model, which is not limited by the embodiment of the present invention. Optionally, a strategy application judgment and optimization suggestion generation model can be set according to experience or actual needs, or can be obtained through model training, which is not limited by the embodiment of the present invention. Exemplarily, a strategy application judgment and optimization suggestion generation model can be obtained by model training through a training strategy application and optimization suggestion generation data set. A training strategy application and optimization suggestion generation data may include but is not limited to a customer service decomposition statement and a strategy application judgment and optimization suggestion label of the corresponding customer service decomposition statement. The number of strategy application judgment and optimization suggestion labels of a customer service decomposition statement can be one or more, and so on; the embodiment of the present invention does not limit this. Optionally, a strategy application judgment and optimization suggestion label may include at least one of the following: a strategy application score and an optimization suggestion, etc., which is not limited by the embodiment of the present invention; optionally, a strategy application judgment and optimization suggestion label can be manually labeled, and so on.

[0076] Optionally, the electronic device may determine the first strategy application judgment and optimization suggestion input data based on the second session data, the first response result and at least one evaluation comparison example pair, and input the first strategy application judgment and optimization suggestion input data into the first strategy application judgment and optimization suggestion generation model, so as to output the first strategy application judgment and optimization suggestion result through the first strategy application judgment and optimization suggestion generation model, thereby calling the first strategy application judgment and optimization suggestion generation model and generating the first strategy application judgment and optimization suggestion result based on the second session data, the first response result and at least one evaluation comparison example pair. Optionally, when determining the first policy application judgment and optimization suggestion input data based on the second session data, the first response result, and at least one evaluation comparison example pair, the electronic device may add the second session data, the first response result, and at least one evaluation comparison example pair to the first policy application judgment and optimization suggestion input data to determine the first policy application judgment and optimization suggestion input data. In this case, the first policy application judgment and optimization suggestion input data may include the second session data, the first response result, and at least one evaluation comparison example pair; or, the second session data, the first response result, the initial policy suggestion, and at least one evaluation comparison example pair may be added to the first policy application judgment and optimization suggestion input data. In this case, the first policy application judgment and optimization suggestion input data may include the second session data, the first response result, the initial policy suggestion, and at least one evaluation comparison example pair; or, the first policy application judgment and optimization suggestion generation prompt information may be determined, and the second session data, the first response result, the initial policy suggestion, at least one evaluation comparison example pair, and the first policy application judgment and optimization suggestion generation prompt information may be added to the first policy application judgment and optimization suggestion input data, and so on; the embodiments of the present invention are not limited to this. Exemplarily, the first strategy application judgment and optimization suggestion input data may also include customer service goals, etc. Optionally, the first strategy application judgment and optimization suggestion generation prompt information may be set according to experience or according to actual needs, and the embodiment of the present invention does not limit this. Exemplarily, assuming that the first strategy application judgment and optimization suggestion input data includes the second session data, the first response result, the initial strategy suggestion, at least one judgment comparison example pair, and the first strategy application judgment and optimization suggestion generation prompt information, then the first strategy application judgment and optimization suggestion input data can be "{{second session data}}{{first response result (here, it means that the current response content is the first response result)}}{{initial strategy suggestion (here, it means that the strategy label is the initial strategy suggestion)}}{{at least one judgment comparison example pair}}".The first strategy application evaluation and optimization suggestion generation prompt information is as follows: Please judge whether the current response content is consistent with the content in the strategy tag based on the conversation context data. The strategy application score is 1-3, where 1 point represents poor strategy application or no relevant suggestions are used, 2 points represent that the strategy suggestions are basically applied, but there are still obvious shortcomings, and 3 points represent that the application is in line with the strategy. Please generate a strategy application score, explain your judgment basis, and give your optimization suggestions, etc. Based on this, the embodiment of the present invention can draw on the Chain-of-Thought (CoT) processing method to guide the strategy application evaluation and optimization suggestion generation model to learn the essence of strategy suggestions by constructing example pairs, and judge the current response result in combination with the conversation context, so as to provide judgment basis and optimization suggestions.

[0077] Optionally, in other embodiments, at least one evaluation and comparison example pair may not be used to generate the first strategy application evaluation and optimization suggestion result, that is, at least one evaluation and comparison example pair may not be used to determine the first strategy application evaluation and optimization suggestion input data; illustratively, the first strategy application evaluation and optimization suggestion input data may include but is not limited to at least one of the following: second session data, first response result, initial strategy recommendation, and first strategy application evaluation and optimization suggestion generation prompt information, etc. The embodiment of the present invention is not limited to this. At this time, the first strategy application evaluation and optimization suggestion result can be generated based on at least one of the second session data, first response result, initial strategy recommendation, and first strategy application evaluation and optimization suggestion generation prompt information, etc.

[0078] S305: Determine the target customer service response content based on the first strategy application evaluation and optimization suggestion result and the first response result, and output the target customer service response content.

[0079] In an embodiment of the present invention, the electronic device may determine whether the first response result meets the response condition based on the first policy application judgment and optimization suggestion result. Optionally, the response condition may be set according to experience or according to actual needs, and the embodiment of the present invention does not limit this; illustratively, the response condition may refer to the response score (i.e., the response score corresponding to the corresponding policy application judgment and optimization suggestion result) reaching a preset response score threshold or the number of iterative generation times of the response result reaching a preset response result iterative generation number threshold, and so on. Optionally, both the preset response score threshold and the preset response result iterative generation number threshold may be set according to experience or according to actual needs, and the embodiment of the present invention does not limit this. Optionally, when the number of policy application scores in a policy application judgment and optimization suggestion result is one, the policy application score in the corresponding policy application judgment and optimization suggestion result may be used as the response score; when the number of policy application scores in a policy application judgment and optimization suggestion result is multiple, the average between the multiple policy application scores in the corresponding policy application judgment and optimization suggestion result may be used as the response score, and so on; the embodiment of the present invention does not limit this. In an embodiment of the present invention, the number of iterative generation times of a response result may refer to the number of iterations for generating a response result for a message to be replied to, such as the first response result may be the response result generated for the current message to be replied to (i.e., the message to be replied to entered by the target customer) when the number of iterative generation times of the response result is 1. Optionally, when the response condition refers to the response score reaching a preset response score threshold or the number of iterative generation times of the response result reaching a preset threshold for iterative generation times of the response result, the response score may be determined by the current strategy application judgment and optimization suggestion result (such as the first strategy application judgment and optimization suggestion result). If the response score reaches the preset response score threshold, it may be determined that the current response result (such as the first response result) meets the response condition, or if the number of iterative generation times of the response result reaches the preset threshold for iterative generation times of the response result, it may be determined that the current response result meets the response condition, and so on.

[0080] Based on this, when the first response result meets the response conditions, the first response result can be used as the target customer service response content; when the first response result does not meet the response conditions, a second response result is generated based on the first session data and the first strategy application evaluation and optimization suggestion results (that is, the response result generated when the response result is iterated 2 times), and a second strategy application evaluation and optimization suggestion result is generated based on the second session data and the second response result, so as to judge whether the second response result meets the response conditions based on the second strategy application evaluation and optimization suggestion results, until the current response result meets the response conditions, so that the current response result can be used as the target customer service response content.

[0081] Optionally, when generating a second response result based on the first session data and the first policy application judgment and optimization suggestion results, the electronic device may call the second response result generation model to generate a second response result based on the first session data and the first policy application judgment and optimization suggestion results; that is, the electronic device may determine the second response result generation model input data based on the first session data and the first policy application judgment and optimization suggestion results, and input the second response result generation model input data into the second response result generation model to output the second response result through the second response result generation model. Based on this, the second response result generation model input data may include the first session data and the first policy application judgment and optimization suggestion results; optionally, the second response result generation model input data may also include but is not limited to at least one of the following: initial policy recommendations, customer service goals, and second preset result generation prompt information, etc., which is not limited in the embodiment of the present invention. Optionally, the second preset result generation prompt information may be set according to experience or according to actual needs, which is not limited in the embodiment of the present invention. Based on this, the embodiment of the present invention may combine the first policy application judgment and optimization suggestion results to regenerate a second response result that is more in line with the initial policy recommendation, such as Figure 2 As shown by arrow 6 in FIG.

[0082] Optionally, when generating a second policy application evaluation and optimization suggestion result based on the second session data and the second response result, the electronic device may invoke a second policy application evaluation and optimization suggestion generation model to generate the second policy application evaluation and optimization suggestion result based on the second session data and the second response result. Accordingly, the electronic device may determine second policy application evaluation and optimization suggestion input data based on the second session data and the second response result, and input the second policy application evaluation and optimization suggestion input data into the second policy application evaluation and optimization suggestion generation model, so that the second policy application evaluation and optimization suggestion result is outputted by the second policy application evaluation and optimization suggestion generation model. In this regard, the second policy application evaluation and optimization suggestion input data may include the second session data and the second response result. Optionally, the second policy application evaluation and optimization suggestion input data may also include, but is not limited to, at least one of the following: at least one evaluation comparison example pair, an initial policy recommendation, a first policy application evaluation and optimization suggestion result, prompt information for generating the second policy application evaluation and optimization suggestion, and a customer service goal, etc., which is not limited in this embodiment of the present invention. Optionally, the prompt information for generating the second policy application evaluation and optimization suggestion may be set based on experience or actual needs, which is not limited in this embodiment of the present invention. Optionally, the second strategy application evaluation and optimization suggestion generation model may be the same as or different from the first strategy application evaluation and optimization suggestion generation model, and the embodiment of the present invention does not limit this. Based on this, the embodiment of the present invention can make another evaluation to obtain the second strategy application evaluation and optimization suggestion result; for example, Figure 2 As shown by arrows 7, 8 and 9 in FIG.

[0083] Optionally, in subsequent iterations, when the second response result does not meet the response conditions, the second response result generation model can be iteratively called to generate the current response result, and the second strategy application evaluation and optimization suggestion generation model can be called to generate the current strategy application evaluation and optimization suggestion result, until the current response result is determined to meet the response conditions based on the current strategy application evaluation and optimization suggestion result, and so on. Optionally, in the subsequent iteration process (i.e., when the second response result does not meet the response condition), the input data of the second response result generation model may include but is not limited to at least one of the following: first session data, at least one strategy application judgment and optimization suggestion result (such as all strategy application judgment and optimization suggestion results that have been generated or the strategy application judgment and optimization suggestion results generated in the previous iteration process, etc.), initial strategy recommendation, customer service goal, and second preset result generation prompt information, etc., and the embodiment of the present invention is not limited to this; Optionally, the input data of the second strategy application judgment and optimization suggestion generation model may include but is not limited to at least one of the following: second session data, current response result, at least one judgment comparison example pair, initial strategy recommendation, at least one strategy application judgment and optimization suggestion result (such as all strategy application judgment and optimization suggestion results that have been generated or the strategy application judgment and optimization suggestion results generated in the previous iteration process, etc.), second strategy application judgment and optimization suggestion generation prompt information, and customer service goal, etc. The embodiment of the present invention is not limited to this.

[0084] Optionally, in other embodiments, the electronic device may also directly use the first response result as the target customer service response content, and so on; the present invention is not limited to this.

[0085] Optionally, in subsequent rounds of conversations, that is, after the target customer service inputs new information to be replied, the electronic device can re-give corresponding initial strategy suggestions for the new information to be replied entered by the new round of target customers based on historical context information, and each initial strategy suggestion has a corresponding explanation, thereby continuously using the customer service response method proposed in the embodiment of the present invention to generate corresponding target customer service response content.

[0086] Optionally, the electronic device may include a policy recommendation component and a policy evaluation component. In this case, the electronic device may generate an initial policy recommendation through the policy recommendation component, and generate any policy application evaluation and optimization recommendation results through the policy evaluation component, and so on.

[0087] Optionally, the customer service response method proposed in the embodiment of the present invention can be applied to any platform, such as an e-commerce shopping platform; illustratively, the customer service response method is applied to an e-commerce shopping platform as an example. The customer system may face many different scenarios, such as pre-sales product consultation, product tracking during sales, and post-sales product returns and exchanges, etc., and each stage may face very different specific scenarios, and each stage can formulate different response goals (i.e., customer service goals) in combination with specific business, and so on.

[0088] From the above, it can be seen that the embodiments of the present invention can enhance the "people-oriented" service concept, that is, the embodiments of the present invention can integrate the knowledge of expert fields into the large language model by introducing communication and emotion regulation skills, thereby achieving more humanized interaction and improving customer satisfaction; and, before responding, the embodiments of the present invention can make initial strategy suggestions for the current round of responses based on the contextual information in the communication process, and make more reasonable speech responses from the perspective of communication skills and methods. In order to achieve the best use of the strategy, the embodiments of the present invention can score the use of the strategy and make optimization suggestions after the response result is generated, thereby obtaining better target customer service response content, etc. In addition, the embodiments of the present invention can construct a corresponding customer service system through the customer service response method as an independent system, so that it can be adapted to any dialogue system and applied to any platform, thereby improving customer experience, etc.

[0089] When a customer service response requirement is detected, an embodiment of the present invention can obtain first session data, where the first session data includes the information to be replied to entered by the target customer. Then, an initial strategy recommendation can be generated based on the first session data, and a first response result can be generated based on the first session data and the initial strategy recommendation. Furthermore, the second session data can be determined, and at least one evaluation comparison example pair can be determined, where a evaluation comparison example pair includes an evaluation response content and evaluation recommendation information for the corresponding evaluation response content; thereby calling the first strategy application evaluation and optimization recommendation generation model, and generating the first strategy application evaluation and optimization recommendation result based on the second session data, the first response result, and at least one evaluation comparison example pair. Based on this, the target customer service response content can be determined based on the first strategy application evaluation and optimization recommendation result and the first response result, and the target customer service response content can be output. It can be seen that the embodiment of the present invention proposes a complete customer service response strategy construction method, which can make better target customer service response content through the strategy recommendation component and the strategy evaluation optimization component (i.e., the strategy application evaluation and optimization recommendation component), and can be applied to various task scenarios. Even in complex business scenarios, it can also assist manual customer service to provide rational and clear replies; based on this, the embodiment of the present invention can flexibly determine the target customer service response content, effectively avoid mechanized customer service responses, thereby effectively improving the response effect of the target customer service response content (i.e., the response effect of the customer service response can be improved by the target customer service response content), thereby effectively improving the customer experience.

[0090] Based on the description of the relevant embodiments of the above-mentioned customer service response method, the embodiment of the present invention further proposes a customer service response device, which can be a computer program (including program code) running in an electronic device; Figure 4 As shown, the customer service response device may include an acquisition unit 401 and a processing unit 402. The customer service response device may execute Figure 1 or Figure 3 The customer service response method shown, that is, the customer service response device can run the above units:

[0091] The acquisition unit 401 is configured to acquire first session data when a customer service response requirement is detected, wherein the first session data includes information to be replied to input by the target customer;

[0092] The processing unit 402 is configured to generate an initial policy suggestion based on the first session data, and generate a first response result based on the first session data and the initial policy suggestion;

[0093] The processing unit 402 is further configured to determine second session data, and generate a first policy application evaluation and optimization recommendation result based on the second session data and the first response result, wherein the second session data includes the information to be replied;

[0094] The processing unit 402 is further configured to determine target customer service response content based on the first strategy application evaluation and optimization suggestion result and the first response result, and output the target customer service response content.

[0095] In one embodiment, the initial policy suggestion is generated based on a policy suggestion library, where the policy suggestion library includes multiple policy suggestions, each policy suggestion including a policy suggestion identifier and an explanation of the corresponding policy suggestion. When generating the initial policy suggestion based on the first session data, the processing unit 402 may be specifically configured to:

[0096] Invoking a first policy suggestion generation model to generate, based on the first session data, a policy suggestion probability value for each policy suggestion under the first session data;

[0097] Based on the policy recommendation probability values of each policy recommendation under the first session data, the top M policy recommendations with the largest policy recommendation probability values are determined from the policy recommendation library, and the M policy recommendations are used as initial policy recommendations to generate initial policy recommendations based on the first session data, where M is a positive integer.

[0098] In another embodiment, the acquisition unit 401 can also be used to

[0099] Acquire third session data, where the third session data includes customer service data;

[0100] The processing unit 402 may also be configured to:

[0101] Invoking the second strategy suggestion generation model, based on the third session data, to determine a label and an explanation for each customer service disassembled statement in the customer service data, to obtain a plurality of label tags and an explanation for each of the plurality of label tags, wherein the label of a customer service disassembled statement is one of the plurality of label tags, and the explanation of a customer service disassembled statement is the explanation of the label of the corresponding customer service disassembled statement;

[0102] The strategy suggestion library is constructed based on the multiple annotation tags and the explanations of the respective annotation tags.

[0103] In another embodiment, when constructing the policy suggestion library based on the multiple annotation tags and the explanations of the respective annotation tags, the processing unit 402 may be specifically configured to:

[0104] Performing vector representation on each of the plurality of annotation labels to obtain a label embedding vector for each annotation label, and determining a target embedding vector corresponding to the corresponding annotation label based on the label embedding vector for each annotation label;

[0105] performing label filtering on the plurality of label tags based on target embedding vectors corresponding to the respective label tags to obtain at least one filtered label tag, and determining an explanation of each filtered label tag in the at least one filtered label tag from the explanations of the respective label tags;

[0106] The at least one screening annotation tag and the explanations of the respective screening annotation tags are used to construct the policy suggestion library, so that a screening annotation tag and the explanation of the corresponding screening annotation tag serve as a policy suggestion in the policy suggestion library.

[0107] In another embodiment, when the processing unit 402 generates the first policy application evaluation and optimization suggestion result based on the second session data and the first response result, it can be specifically configured to:

[0108] Determining at least one evaluation comparison example pair, where each evaluation comparison example pair includes an evaluation response content and evaluation suggestion information corresponding to the evaluation response content;

[0109] The first strategy application evaluation and optimization suggestion generation model is called to generate a first strategy application evaluation and optimization suggestion result based on the second session data, the first response result, and the at least one evaluation and comparison example pair.

[0110] In another implementation, the acquiring unit 401 is further configured to:

[0111] Obtain multiple evaluation response contents corresponding to each of the multiple strategy suggestions;

[0112] The processing unit 402 may also be configured to:

[0113] Determine the evaluation suggestion information of each evaluation response content corresponding to each of the strategy suggestions respectively, and construct a plurality of evaluation comparison example pairs corresponding to each of the strategy suggestions based on each evaluation response content corresponding to each of the strategy suggestions and the evaluation suggestion information of each evaluation response content corresponding to each of the strategy suggestions, wherein a evaluation comparison example pair corresponding to each strategy suggestion includes a evaluation response content corresponding to the corresponding strategy suggestion and the evaluation suggestion information of the corresponding evaluation response content;

[0114] When determining at least one evaluation comparison example pair, the processing unit 402 may be specifically configured to:

[0115] At least one evaluation and comparison example pair is determined from the multiple evaluation and comparison example pairs corresponding to the respective strategy suggestions.

[0116] In another embodiment, when determining the target customer service response content based on the first strategy application evaluation and optimization suggestion result and the first response result, the processing unit 402 may be specifically configured to:

[0117] Determining whether the first response result meets a response condition based on the first strategy application evaluation and optimization suggestion result;

[0118] When the first response result meets the response condition, the first response result is used as the target customer service response content;

[0119] When the first response result does not meet the response condition, a second response result is generated based on the first session data and the first strategy application evaluation and optimization suggestion result, and a second strategy application evaluation and optimization suggestion result is generated based on the second session data and the second response result, so as to judge whether the second response result meets the response condition based on the second strategy application evaluation and optimization suggestion result, until the current response result meets the response condition, so as to use the current response result as the target customer service response content.

[0120] According to one embodiment of the present invention, Figure 4 Each unit in the customer service answering device shown can be individually or completely combined into one or several other units to form a structure, or one (or some) of the units can be further divided into multiple functionally smaller units to form a structure, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present invention. The above-mentioned units are divided based on logical functions. In actual applications, the functions of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present invention, any customer service answering device can also include other units. In actual applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0121] According to another embodiment of the present invention, the program can be executed by running a program on a general electronic device such as a computer including a central processing unit (CPU), a random access memory (RAM), a read-only memory (ROM), and other processing elements and storage elements. Figure 1 or Figure 3 A computer program (including program code) for each step involved in the corresponding method shown in Figure 4 The customer service response device shown in and the customer service response method of the embodiment of the present invention are implemented. The computer program can be recorded on, for example, a computer storage medium, and loaded into the above-mentioned electronic device through the computer storage medium and run therein.

[0122] When a customer service response request is detected, embodiments of the present invention can obtain first session data, the first session data including the information to be replied to entered by the target customer; generate an initial strategy recommendation based on the first session data, and generate a first response result based on the first session data and the initial strategy recommendation. Then, second session data can be determined, and based on the second session data and the first response result, a first strategy application evaluation and optimization recommendation result can be generated, the second session data including the information to be replied to. Furthermore, the target customer service response content can be determined based on the first strategy application evaluation and optimization recommendation result and the first response result, and the target customer service response content can be output. It can be seen that embodiments of the present invention can achieve the comprehensive improvement of the intelligence and humanization of customer service responses by introducing communication and emotion regulation skills through the initial strategy recommendation by generating an initial strategy recommendation and introducing the initial strategy recommendation into the response result generation process. This allows for the flexible determination of the target customer service response content, effectively improving the response effectiveness of the target customer service response content, and thereby effectively enhancing the customer experience.

[0123] Based on the description of the above method embodiment and apparatus embodiment, the exemplary embodiments of the present invention further provide an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, and when executed by the at least one processor, the computer program causes the electronic device to perform a method according to an embodiment of the present invention.

[0124] Exemplary embodiments of the present invention further provide a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to perform a method according to an embodiment of the present invention.

[0125] An exemplary embodiment of the present invention further provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor of a computer, the computer is configured to cause the computer to perform a method according to an embodiment of the present invention.

[0126] refer to Figure 5 , a block diagram of an electronic device 500 that can serve as a server or client of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0127] like Figure 5 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. Various programs and data required for the operation of the electronic device 500 can also be stored in the RAM 503. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0128] Multiple components within electronic device 500 are connected to I / O interface 505, including an input unit 506, an output unit 507, a storage unit 508, and a communication unit 509. Input unit 506 can be any type of device capable of inputting information into electronic device 500. Input unit 506 can receive input numeric or character information and generate key input signals related to user settings and / or function control of the electronic device. Output unit 507 can be any type of device capable of presenting information and may include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. Storage unit 508 may include, but is not limited to, a magnetic disk or an optical disk. Communication unit 509 allows electronic device 500 to exchange information / data with other devices via computer networks such as the Internet and / or various telecommunication networks and may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver and / or chipset, such as a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0129] The computing unit 501 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above. For example, in some embodiments, the customer service response method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 500 via the ROM 502 and / or the communication unit 509. In some embodiments, the computing unit 501 can be configured to execute the customer service response method in any other appropriate manner (e.g., by means of firmware).

[0130] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0131] In the context of the present invention, machine-readable medium can be a tangible medium that can contain or store a program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0132] As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device (e.g., a magnetic disk, an optical disk, a memory, a programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0133] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0134] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0135] Computer systems may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The client and server relationship arises through computer programs running on the respective computers and having a client-server relationship to each other.

[0136] Furthermore, it should be understood that the above disclosure is only a preferred embodiment of the present invention and certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A customer service response method, characterized in that: include: When a customer service response requirement is detected, first session data is obtained, where the first session data includes information to be replied to entered by the target customer; generating an initial policy suggestion based on the first session data, and generating a first response result based on the first session data and the initial policy suggestion; Determining second session data, and generating a first strategy application evaluation and optimization recommendation result based on the second session data and the first response result, wherein the second session data includes the information to be replied; Based on the first strategy application evaluation and optimization suggestion result and the first response result, target customer service response content is determined, and the target customer service response content is output.

2. The method according to claim 1, characterized in that The initial policy suggestion is generated based on a policy suggestion library, wherein the policy suggestion library includes multiple policy suggestions, and each policy suggestion includes a policy suggestion identifier and an explanation of the corresponding policy suggestion. Generating the initial policy suggestion based on the first session data includes: Invoking a first policy suggestion generation model to generate, based on the first session data, a policy suggestion probability value for each policy suggestion under the first session data; Based on the policy recommendation probability values of each policy recommendation under the first session data, the top M policy recommendations with the largest policy recommendation probability values are determined from the policy recommendation library, and the M policy recommendations are used as initial policy recommendations to generate initial policy recommendations based on the first session data, where M is a positive integer.

3. The method according to claim 2, characterized in that The method further comprises: Acquire third session data, where the third session data includes customer service data; Invoking the second strategy suggestion generation model, based on the third session data, to determine a label and an explanation for each customer service disassembled statement in the customer service data, to obtain a plurality of label tags and an explanation for each of the plurality of label tags, wherein the label of a customer service disassembled statement is one of the plurality of label tags, and the explanation of a customer service disassembled statement is the explanation of the label of the corresponding customer service disassembled statement; The strategy suggestion library is constructed based on the multiple annotation tags and the explanations of the respective annotation tags.

4. The method according to claim 3, characterized in that The step of constructing the strategy suggestion library based on the multiple annotation tags and the explanations of the respective annotation tags includes: Performing vector representation on each of the plurality of annotation labels to obtain a label embedding vector for each annotation label, and determining a target embedding vector corresponding to the corresponding annotation label based on the label embedding vector for each annotation label; performing label filtering on the plurality of label tags based on target embedding vectors corresponding to the respective label tags to obtain at least one filtered label tag, and determining an explanation of each filtered label tag in the at least one filtered label tag from the explanations of the respective label tags; The at least one screening annotation tag and the explanations of the respective screening annotation tags are used to construct the policy suggestion library, so that a screening annotation tag and the explanation of the corresponding screening annotation tag serve as a policy suggestion in the policy suggestion library.

5. The method according to any one of claims 1 to 4, characterized in that The generating, based on the second session data and the first response result, a first strategy application evaluation and optimization suggestion result includes: Determining at least one evaluation comparison example pair, where each evaluation comparison example pair includes an evaluation response content and evaluation suggestion information corresponding to the evaluation response content; The first strategy application evaluation and optimization suggestion generation model is called to generate a first strategy application evaluation and optimization suggestion result based on the second session data, the first response result, and the at least one evaluation and comparison example pair.

6. The method according to claim 5, characterized in that The method further comprises: Obtain multiple evaluation response contents corresponding to each of the multiple strategy suggestions; Determine the evaluation suggestion information of each evaluation response content corresponding to each of the strategy suggestions respectively, and construct a plurality of evaluation comparison example pairs corresponding to each of the strategy suggestions based on each evaluation response content corresponding to each of the strategy suggestions and the evaluation suggestion information of each evaluation response content corresponding to each of the strategy suggestions, wherein a evaluation comparison example pair corresponding to each strategy suggestion includes a evaluation response content corresponding to the corresponding strategy suggestion and the evaluation suggestion information of the corresponding evaluation response content; The determining of at least one comparison example pair includes: At least one evaluation and comparison example pair is determined from the multiple evaluation and comparison example pairs corresponding to the respective strategy suggestions.

7. The method according to any one of claims 1 to 4, characterized in that The step of determining the target customer service response content based on the first strategy application evaluation and optimization suggestion result and the first response result includes: Determining whether the first response result meets a response condition based on the first strategy application evaluation and optimization suggestion result; When the first response result meets the response condition, the first response result is used as the target customer service response content; When the first response result does not meet the response condition, a second response result is generated based on the first session data and the first strategy application evaluation and optimization suggestion result, and a second strategy application evaluation and optimization suggestion result is generated based on the second session data and the second response result, so as to judge whether the second response result meets the response condition based on the second strategy application evaluation and optimization suggestion result, until the current response result meets the response condition, so as to use the current response result as the target customer service response content.

8. A customer service answering device, characterized in that: The device comprises: an acquiring unit, configured to acquire first session data when a customer service response requirement is detected, wherein the first session data includes information to be replied to input by the target customer; a processing unit, configured to generate an initial policy suggestion based on the first session data, and generate a first response result based on the first session data and the initial policy suggestion; The processing unit is further configured to determine second session data, and generate a first policy application evaluation and optimization recommendation result based on the second session data and the first response result, wherein the second session data includes the information to be replied; The processing unit is further configured to determine target customer service response content based on the first strategy application evaluation and optimization suggestion result and the first response result, and output the target customer service response content.

9. An electronic device, characterized in that: include: processor; as well as Memory for storing programs, The program includes instructions, which, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images