Customer service information processing method, device and system, electronic equipment and storage medium

Abstract

The invention discloses a customer service information processing method, device and system, electronic equipment and a storage medium. The method applied to a server comprises the following steps: when a customer service request from a user side is received, obtaining text data of user consultation content from the customer service request; inputting the text data of the user consultation content into a large model obtained by fine tuning in advance and a classification model trained in advance, obtaining primary reply content for the user consultation content through the large model, and obtaining a service type matched with the user consultation content through the classification model; generating an auditing task based on the user consultation content and the primary reply content thereof, and sending the auditing task to a customer service end of the same service type; and receiving reviewed reply content returned by the customer service side, and sending the reviewed reply content to the user side sending the customer service request. By utilizing the embodiment disclosed by the invention, the working efficiency of customer service can be improved, and the controllability of reply contents is ensured.

Description

Technical Field

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

[0002] With the development of Internet application (Application, APP for short) technology and portal website technology, more and more enterprises, institutions or individuals, etc. realize corresponding business online through APP, portal websites, etc., and the business content will also change and adjust over time. In the process of business implementation, in order to better serve the business-oriented users, customer service (referred to as customer service) such as online business Q&A and online troubleshooting is usually provided. When the usual customer service system receives the consulting questions input by the user through the user terminal, it transmits the consulting questions to the staff, who manually inputs the reply content, and the customer service system returns the reply content to the user terminal. When the number of users is large and the number of consulting questions is large, this manual customer service system requires a large number of staff, so the labor cost is high.

[0003] With the development of artificial intelligence (AI) technology, more and more customer systems use AI technology to answer questions raised by users. For example, some online businesses in the financial industry use intelligent customer service question-and-answer systems to link to knowledge bases, provide answers to routine, repetitive simple questions, and retrieve answers to user questions in the knowledge base through classification lists, keywords / words, and configured regular expressions. For another example, the Chinese patent invention with publication number CN 113918697 A and invention name "An optimization method and optimization system for an intelligent question-and-answer system" uses a BERT learning model to convert user questions into vectors, and uses the Milvus vector search engine to search in the database and obtain the recall data with the highest similarity to the vector.

[0004] The customer service system using AI technology directly returns the answer to the user's question after obtaining the answer to the user's question. The reply content is often uncontrollable, resulting in the user receiving a reply that does not conform to the customer service concept or even goes against the customer service concept. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a customer service information processing method, device, system, electronic device and storage medium, which improves the work efficiency of customer service personnel while ensuring the controllability of reply content.

[0006] According to one aspect of the present invention, an embodiment of the present invention provides a customer service information processing method, which is applied to a server and includes the following steps:

[0007] When receiving a customer service request from a user, obtaining text data of the user's consultation content from the customer service request;

[0008] The text data of the user's consultation content is input into the pre-fine-tuned large model and the pre-trained classification model respectively, the large model obtains the primary reply content for the user's consultation content, and the classification model obtains the business type matching the user's consultation content;

[0009] Generate an audit task based on the user consultation content and its primary reply content, and store the user consultation content and its primary reply content in the audit task into an audit task pool corresponding to the business type;

[0010] Sending audit tasks in the audit task pool to the client end of the same business type based on the task allocation strategy; and

[0011] Receive the review completion reply content returned by the customer service end, and send the review completion reply content to the user end that issued the customer service request.

[0012] According to another aspect of the present invention, an embodiment of the present invention provides a customer service information processing device, including a consulting content acquisition module, a large model module, a classification model module, an audit task generation module, an audit task allocation module and a reply module, wherein the consulting content acquisition module is configured to receive a customer service request from a user end, and obtain text data of the user consulting content from the customer service request; the large model module is configured to input the text data of the user consulting content into the pre-fine-tuned large model, and obtain the primary reply content for the user consulting content from the large model; the classification model module is configured to input the text data of the user consulting content into a pre-trained classification model, and obtain the business type matching the user consulting content from the classification model; the audit task generation module is configured to generate an audit task based on the user consulting content and its primary reply content, and store the user consulting content and its primary reply content in the audit task into an audit task pool corresponding to the business type; the audit task allocation module is configured to send the audit task in the audit task pool to the customer service end of the same business type based on the task allocation strategy; the reply module is configured to receive the reply content of the audit completed returned by the customer service end, and send the reply content of the audit completed to the user end that issued the customer service request.

[0013] According to another aspect of the present invention, an embodiment of the present invention provides a customer service information processing system, including a user terminal device, a customer service information processing device and a customer service terminal device, wherein the user terminal device is configured as a user end, and after receiving the consultation content and instructions input by the user, sends a customer service request to the service end, and includes the user consultation content in the customer service request, and receives the returned reply content; the customer service information processing device acts as a service end, communicates with the user terminal device through a network, and is configured to obtain the primary reply content corresponding to the user consultation content based on the large model obtained by pre-fine-tuning when receiving the customer service request sent by the user terminal device, obtain the business type matching the user consultation content based on the classification model, generate an audit task based on the user consultation content and its primary reply content, and store the user consultation content and its primary reply content in the audit task in the audit task pool corresponding to the business type; send the audit task in the audit task pool to the customer service end of the same business type based on the task allocation strategy; and send the reply content reviewed by the customer service end to the user terminal device that issued the customer service request; the customer service terminal device is configured as a customer service end, receives the audit task and sends the audited reply content to the customer service information processing device after the customer service personnel completes the audit.

[0014] According to another aspect of the present invention, an embodiment of the present invention provides an electronic device, which includes a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the aforementioned customer service information processing method is implemented.

[0015] According to another aspect of the present invention, an embodiment of the present invention provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the aforementioned customer service information processing method is implemented.

[0016] According to another aspect of the present invention, an embodiment of the present invention provides a computer program product, which includes computer program instructions, and the computer program instructions implement the aforementioned customer service information processing method when executed by a processor.

[0017] The present invention provides preliminary responses to consulting questions through a large model, thereby improving the work efficiency of customer service personnel, and the response content of the large model needs to be assigned to customer service personnel for review, thereby ensuring the controllability of the response content. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the following briefly introduces the drawings in the embodiment of the present invention.

[0019] Figure 1It is a principle framework diagram of a customer service information processing system applied to a business system according to an embodiment of the present invention.

[0020] Figure 2 It is a flow chart of a customer service information processing method according to an embodiment of the present invention.

[0021] Figure 3 It is a principle block diagram of a customer service information processing device according to an embodiment of the present invention.

[0022] Figure 4 It is a principle block diagram of the large model module 12 and the classification model module 13 according to an embodiment of the present invention.

[0023] Figure 5 FIG. 4 is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that the purpose of providing these embodiments is to make the principles and spirit of the present invention clearer and more thorough, so that those skilled in the art can better understand and implement the principles and spirit of the present invention. The exemplary embodiments provided herein are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments herein, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method, computer-readable storage medium, or computer program product. Therefore, the present disclosure may be implemented in at least one of the following forms: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0026] According to the embodiments of the present invention, the present invention claims protection for a device, an electronic device, a computer-readable storage medium, and a computer program product.

[0027] In this document, terms such as first, second, etc. are only used to distinguish one entity (or operation) from another entity (or operation), and do not require or imply any order or association between these entities (or operations). In this document, the elements (such as parts, components, processes, steps) defined by the sentence "including..." do not exclude the existence of other elements in addition to the listed elements, that is, other elements that are not explicitly listed may also be included. In this document, any elements and their quantities in the drawings are used for illustration rather than limitation, and any names in the drawings are only used for distinction and do not have any limiting meaning.

[0028] The present invention relates to a customer service information processing method, device and system applied to a business system. The business system can realize a single business or multiple different businesses at the same time. For example, a video business system includes both video push business and live broadcast business, and of course other businesses, such as advertising business, various promotional activities business, etc. In order to provide customer service to users, the business system also includes a customer service system, such as Figure 1 As shown, Figure 1 1 is a principle framework diagram of a customer service information processing system applied to a business system according to an embodiment of the present invention. The customer service information processing system includes a user side and a service side, and the user side interacts with the service side through a network. The service side includes a first server 1041 and a customer service terminal device, wherein the customer service terminal device includes a first customer service terminal device 1061 connected to the first server 1041 through an internal communication channel, and may also include a second customer service terminal device 1062 connected to the first server 1041 through an external network. The first server 1041 and the second customer service terminal device 1062 may also receive and send messages through a third-party message channel, wherein the second server 1042 is a third-party server, and is used for message transmission between the first server 1041 and the second customer service terminal device 1062. Among them, the user terminal device 102 and the customer service terminal device may be at least one of the following devices: a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart TV, and the like. An application client is installed on the user terminal device 102. The application client can be a client that provides business-related services (such as an application APP), or a client with an embedded application applet (different functions). It can also be a client that logs in through a browser and inputs instructions through client operations. Alternatively, the user can open the browser installed on the user terminal device 102 and input instructions through browser operations. The client includes a customer service module (such as the client software used by users in the customer service system), and users can input consulting content through the customer service module. After the user terminal device 102 receives the instruction entered by the user, it sends the request information containing the instruction to the first server 1041, which serves as the business service end. Although Figure 1 Only one first server 1041 is shown in the figure. Those skilled in the art should understand that there may be multiple servers on the service end, such as a server for business processing, a server for user information management, a server for data management, etc. After receiving the request from the client, the first server 1041 performs corresponding processing to obtain the primary reply content and generates an audit task, and assigns the audit task to a specific customer service staff. The customer service staff conducts the audit through the customer service terminal device and sends the audit result to the first server 1041. The first server 1041 sends the audited reply content to the user terminal device 102.

[0029] The customer service information processing method, device and system of the present invention are described in detail below through specific embodiments.

[0030] Figure 2 FIG. 1 is a flow chart of a method for processing customer service information according to an embodiment of the present invention. In this embodiment, upon receiving a customer service request from a user, the following processing steps are included:

[0031] Step S11, obtaining text data of the user's consultation content from the customer service request.

[0032] Step S12, inputting the text data of the user's inquiry content into the pre-fine-tuned large model, and obtaining the primary reply content for the user's inquiry content from the large model.

[0033] Step S13, inputting the text data of the user consultation content into a pre-trained classification model, and obtaining the business type matching the user consultation content from the classification model.

[0034] Step S14, generating an audit task based on the user consultation content and its primary reply content, and storing the user consultation content and its primary reply content in the audit task into an audit task pool corresponding to the business type.

[0035] Step S15: sending the audit tasks in the audit task pool to designated client terminals of the same business type based on the task allocation strategy.

[0036] Step S16, receiving the reply content of the review completion returned by the designated customer service end, and sending the reply content of the review completion to the user end that issued the customer service request.

[0037] Among them, see Figure 1, the client installed in the user terminal device 102 includes a user-side customer service module, and the user inputs the consultation content in the form of text, image, voice or recorded video through the user-side customer service module. The client installed in the user terminal device 102 sends the consultation content in text format, image format, audio format or video format to the server. When the server receives the customer service request sent by the client, in step S11, the data in the original format is parsed from the customer service request. If the original format of the data is text format, the text data is input into the pre-fine-tuned large model in step S12. If the original format of the data is audio format, voice recognition is performed to convert the consultation content expressed by voice into text data. If the original format of the data is video format, on the one hand, voice recognition is performed to convert the consultation content expressed by voice into text data, and on the other hand, the video is framed, the corresponding image content is identified from the determined key frame, and the image content is converted into a text description, and the text description is merged into the text data of the consultation content expressed by voice. In addition, the user can also issue a customer service request in the form of text and picture, text and video, and text and voice. In step 11, the picture, voice or video image is also converted into text data, and then merged with the text content in the customer service request as the user's consultation content. The present invention provides users with a variety of consultation methods, so that users can express their questions more clearly in a suitable way, which is more user-friendly, and improves the accuracy of the problem description at the source of the consultation problem, thereby improving the accuracy of the reply, and further improving the customer service efficiency and user satisfaction.

[0038] In step S12, the present invention utilizes the existing open source big model, obtains the big model adapted to the business needs of the present embodiment through pre-fine-tuning, and obtains the primary reply content for the user's inquiry content from the big model.

[0039] When fine-tuning the large model, first obtain historical manual customer service data, process the historical manual customer service data to obtain a training data set, and each training sample in the training data set is a question-answer pair consisting of a consultation content and a reply content. For example:

[0040] {“content”: “Why has the gold coin cash-out decreased?”, “summary”: “Hello, xxx will help you verify the gold coin withdrawal status.”}

[0041] {“content”: “How to download the app on a computer”, “summary”: “Please open the software store / software market that comes with your phone system, search for “XX” and click to download and install”} ......

[0043] In the above question-answer sample, the field "content" is the inquiry content raised by the user, and the field "summary" is the reply content given by the customer service to the inquiry content raised by the user.

[0044] By collecting manual customer service data and organizing it into the question-answer format as above, we can obtain a training dataset for fine-tuning the large model.

[0045] Then, the GLM model is fine-tuned using the training data set based on the P-tuning fine-tuning technology. The P-tuning fine-tuning technology converts the Prompt into a learnable Embedding layer, and uses the MLP+LSTM method to process the PromptEmbedding layer. The specific fine-tuning steps can be found in the paper "GPT Understands, Too", which will not be repeated here.

[0046] After the fine-tuning is completed, a model parameter set that meets the requirements is obtained and stored. Although the present embodiment adopts the GLM large model, it can be understood by those skilled in the art that other models and fine-tuning techniques can also be adopted, and specific references are made to the corresponding model reference guides.

[0047] After step S11 obtains the text data of the consultation content, the GLM large model is called, and the text data is input into the GLM large model, so that the reply content with the input text data as the question can be obtained. Since the present invention finally sends the user the reviewed reply content, which may have been modified during the review process, in order to distinguish it from the final reply content sent to the user, the reply content obtained by the GLM large model is referred to as the primary reply content.

[0048] Usually, in order to reduce the business pressure of customer service personnel and improve the accuracy of responses to relevant business consultations, customer service personnel are classified into different business types. Therefore, in an embodiment of the present invention, a pre-trained classification model is used in step S13 to classify the user consultation content. The classification model is trained using historical manual customer service data as training corpus. The corpus contains questions asked by users and corresponding classifications. The corpus examples are shown in Table 1 below.

[0049] Table 1

[0050]

[0051] Table 1 above is only a part of the corpus, and the classification therein can be determined according to the business needs and the business type of the customer service personnel. For example, it can also include types such as user accounts, advertisements, etc.

[0052] In one embodiment, the classification model uses classification algorithms such as logistic regression. After model training, evaluation, adjustment, and verification, the classification types are stored when the accuracy requirements are met and called when needed.

[0053] After obtaining the primary reply content in step S12, in step S14, an audit task is generated based on the user consultation content and its primary reply content, and the user consultation content and its primary reply content in the audit task are stored in an audit task pool corresponding to the business type determined in step S13.

[0054] In one embodiment, each audit task in the audit task pool has a task label to indicate the status of the audit task, such as "unreviewed", "under review" or "reviewed", etc. In step S14, when the user consultation content and the primary reply content in the audit task are stored in the audit task pool corresponding to the business type, the audit status of the audited task is set to "unreviewed" in the task label of the audit task. In order to facilitate task allocation, optionally, a task ID is generated for each audit task when it is generated. In addition to storing the user consultation content and the primary reply content, other information can also be stored, such as the client ID, which is used to represent the client identification that issued the consultation question.

[0055] In another embodiment, the audit task pool for each business type includes an unaudited task sub-pool and an audited task sub-pool. In step S14, when the user consultation content and the primary reply content in the audit task are stored in the audit task pool corresponding to the business type, the audit task is stored in the unaudited task sub-pool. During the subsequent task allocation, the unaudited task sub-pool is scanned, and an audit task is determined to be sent to a designated client terminal of the same business type; correspondingly, when the reply content of the audit completion returned by the designated client terminal is received, the corresponding audit task is transferred from the unaudited task sub-pool to the audited task sub-pool.

[0056] The server continuously repeats the above steps S11 to S14 to store audit tasks in the audit task pool of each business type. Each audit task in the audit task pool is arranged according to the storage time or task ID.

[0057] In step S15, the task allocation strategy includes a pre-set priority ranking of the client terminals, such as priority ranking in order of the amount of pending tasks of the client terminals from low to high; or priority ranking in order of client terminal IDs from low to high.

[0058] For example, priority is given to assigning tasks to clients with less tasks, or in order to balance the task processing workload of the client, the audit tasks are assigned in the order of the client ID. Among them, the audit tasks to be assigned are determined by scanning the task label of each audit task in the audit task pool, such as determining the audit tasks to be assigned with the audit status of "unaudited" in the task label; or scanning the unaudited task sub-pool to determine one as the audit task to be assigned. When it is determined in the task allocation strategy to allocate in the order of client ID, the determined audit tasks to be assigned are assigned to different client terminals in turn according to the client ID, and after being assigned to the last client, they are returned to the first client for allocation. When it is determined in the task allocation strategy to allocate audit tasks according to the priority of task processing workload, the amount of tasks that have been assigned but not yet processed for each client is obtained, and the client with the lowest current task volume is given priority.

[0059] When the client receives the review task, it checks the user's inquiry content and its primary reply content to determine whether the primary reply content is appropriate. If it is appropriate, it returns the reviewed reply content to the server. If it is not appropriate, it will be manually modified and the modified text will be returned to the server as the reviewed reply content.

[0060] In step S16, when the reply content of the review completion returned by the client is received, it is sent to the user end, and then the review status in the task label of each review task in the review task pool is set to "reviewed"; or the review task is transferred from the unreviewed task sub-pool to the reviewed task sub-pool.

[0061] In one embodiment, when the reply content is reviewed, the reply content modified by the customer service staff and the user's consultation content are collected and used as the corpus or sample data of the large model training set, so that the large model can be trained and updated based on the new training set. The continuous accumulation of corpus improves the reasoning ability of the large model, thereby improving the accuracy of the reply content obtained by the large model.

[0062] In one embodiment, before the reply content after the review is sent to the user end that issued the customer service request, the data format of the user consultation content in the customer service request is also checked; since the reply content after the review in this embodiment is in text format, if it is different from the data format of the user consultation content in the customer service request, the data format of the reply content after the review is converted to the same data format as the data format of the user consultation content in the customer service request. For example, when the data format of the user consultation content in the customer service request is audio, the reply content in text format is converted to voice format. In a better embodiment, in order to improve the credibility of the reply content, different business types correspond to different audio parameters, and audio parameters include, for example, pitch, speech speed, etc. For example, the audio of the APP application download category uses a young male voice, the audio of the live broadcast category uses a female voice, the audio of the video upload category uses a baritone, and the audio of the children's category uses a child's voice.

[0063] When the data format of the user consultation content in the customer service request is in video format, the text-formatted reply content is converted into voice data, and the image data of the customer service personnel whose image matches the business type is determined, and the image data and audio data are processed according to the preset video parameters to obtain video reply data. In a simpler implementation, a video recorded by the customer service personnel of the business type is used as video image data, and the audio data is merged into the video to obtain video reply data.

[0064] In another embodiment, a virtual digital human whose image matches the business type is rendered to obtain the image data, and the image data and audio data of the virtual digital human are processed according to preset video parameters to obtain video reply data. In one embodiment, each business type has a virtual digital human with a corresponding image, and the image includes, for example, the appearance, clothing, hairstyle, etc. of the virtual human. The embodiment of the present invention can improve the user's trust in the reply content from the perspective of sound and vision.

[0065] In an optional embodiment, after sending the audit task in the audit task pool to the designated client end in step S15, the processing time of the audit task by the designated client end is monitored. For example, while sending the audit task in the audit task pool to the designated client end, timing is performed, and the system presets a timing cycle, wherein the timing cycle can be modified by the staff with authority, and timing cycles of different lengths can also be set according to different business types. When the reply content of the audit completion of the client end is not received when the timing cycle is reached, that is, the processing time of the audit task by the client end exceeds the threshold, in order not to affect the user experience, the audit task is sent to the secondary client end, and the processing of the audit task by the designated client end is terminated. The secondary client end is, for example, a customer service staff with more experience or a higher level, and the initially determined customer service end is, for example, a junior customer service staff. By monitoring the reply time of the customer service staff and transferring to other customer service staff when the preset time is exceeded, timely feedback on the user's consultation content is guaranteed and feedback delays caused by junior customer service staff due to work experience and other reasons are prevented.

[0066] Among them, the customer service terminal device can be the backend terminal of the server, that is, Figure 1 The first customer service terminal device 1061 in the system has the client software of the customer service system installed therein. The first server 1041 sends the audit task to the client in the first customer service terminal device 1061 through the internal communication channel, and displays it through the dedicated interface of the client. The customer service personnel can view the audit task assigned to them through the dedicated interface and perform audit operations, such as modifying the content of the document and clicking the audit completion button. The customer service terminal device can also be a remote terminal device that communicates with the first server 1041 through the network, such as Figure 1 The second customer service terminal device 1062 in the second customer service terminal device 1062 has the client software of the customer service system installed in it, and the principle and implementation process are the same as those of the first customer service terminal device 1061. Alternatively, the second customer service terminal device 1062 does not need to install the client software of the customer service system, and the first server 1041 and the second customer service terminal device 1062 use a third-party message channel to send and receive audit tasks (such as through Figure 1 For example, the first server 1041 sends the audit task to the second customer service terminal device 1062 via the network and the second server 1042 in the form of SMS or social messages. After receiving the SMS or social messages, the second customer service terminal device 1062 checks the audit task content in the corresponding message interface, and after completing the audit, returns the reply content of the audit completion to the first server 1041 in the original message channel such as SMS or social messages, and then the first server 1041 sends it to the user.

[0067] In the above-mentioned embodiment, when the consultation content sent by the user terminal device 102 is in audio or video format, the server 104 completes the conversion of text data. When returning the reply content to the user terminal, the server 104 also converts the text data into corresponding audio data or video data. It can be understood that the conversion of data format can also be completed by the client in the user terminal device 102. For example, the client provides a variety of consultation content submission methods, such as text, image, voice and video. When the user selects the text method, the client sends the text data input by the user to the server, and when receiving the reply content returned by the server, it is displayed to the user in text mode. When the user selects the voice method, the client performs voice recognition, obtains the corresponding text content, and then sends the text data to the server. When receiving the reply content returned by the server, the text data is converted into voice and played to the user. When the user chooses the video mode, one processing method is to extract the text data of the consultation content in the user's voice when the user's video recording is completed, and send it to the server. When the server returns the reply content, the reply content is converted into audio data, and a virtual digital human is rendered. The audio data and the rendered virtual digital human image data are combined to obtain a video, and the video is played to the user. Another processing method is, for example, to render a virtual digital human and talk with the user in video mode, extract the text data of the consultation content in the user's voice in real time, and send it to the server. The server obtains the reply content and sends it to the client. The client converts the reply content into the voice of the virtual digital human for playback. When the server does not receive the reply content, the virtual digital human uses the preset words to have a voice conversation with the user, so that the user can consult on related issues in a similar way to real-time video.

[0068] On the other hand, the present invention also provides a customer service information processing device, such as Figure 3 As shown, Figure 31 is a principle block diagram of a customer service information processing device according to an embodiment of the present invention. The customer service information processing device 100 in this embodiment includes a consultation content acquisition module 11, a large model module 12, a classification model module 13, an audit task generation module 14, an audit task assignment module 15 and a reply module 16, wherein the customer service information processing device 100 is located at the service end, such as the first server 1041, and communicates with the user terminal device 102 of the user end and the customer service terminal device (such as the first customer service terminal device 1061 and the second customer service terminal device 1062) of the customer end through the communication module of the service end. When the first server 1041 receives the user customer service request sent by the user end, it sends it to the consultation content acquisition module 11 through the communication module. The consultation content acquisition module 11 obtains the text data of the user consultation content from the customer service request and sends it to the large model module 12, the classification model module 13, and the audit task generation module 14 respectively. The large model module 12 inputs the text data of the user consultation content to the large model obtained by pre-fine-tuning, obtains the primary reply content for the user consultation content from the large model, and sends the primary reply content to the audit task generation module 14. The classification model module 13 is connected to the consultation content acquisition module 11, and inputs the text data of the user consultation content into the pre-trained classification model. The classification model obtains the business type that matches the user consultation content and sends it to the audit task generation module 14. The audit task generation module 14 is connected to the consultation content acquisition module 11, the large model module 12 and the classification model module 13 respectively, generates an audit task based on the user consultation content and its primary reply content, and stores the user consultation content and its primary reply content in the audit task into the audit task pool corresponding to the business type. The audit task allocation module 15 sends the audit tasks in the audit task pool to the customer service end of the same business type through the communication module based on the task allocation strategy. The reply module 16 receives the reply content of the audit completed returned by the customer service end through the communication module, and sends the reply content of the audit completed to the user end that issued the customer service request through the communication module.

[0069] In an optional embodiment, when the data format of the user's consultation content is received in audio format, the consultation content acquisition module 11 performs voice recognition and converts the consultation content expressed by voice into text data. If the original format of the data is in video format, on the one hand, the consultation content acquisition module 11 performs voice recognition and converts the consultation content expressed by voice into text data, and on the other hand, the consultation content acquisition module 11 performs frame extraction processing on the video, identifies the corresponding image content from the determined key frames, and converts the image content into text description, and merges the text description into the text data of the consultation content expressed by voice.

[0070] Alternatively, if Figure 4 As shown, Figure 4 It is a principle block diagram of a large model module 12 and a classification model module 13 according to an embodiment of the present invention. The large model module 12 includes a large model fine-tuning unit 121 and an inference unit 122. The large model fine-tuning unit 121 uses P-tuning to fine-tune the GLM large model based on the training data set, and stores the model and its parameter set obtained after fine-tuning. After collecting a certain number of reply contents modified by customer service personnel, these contents are added to the training data set as new samples, and the large model is fine-tuned based on the new training data set and its parameter set is updated. The inference unit 122 calls the large model to obtain the primary reply content based on the text data of the consultation content sent by the consultation content acquisition module 11.

[0071] The classification model module 13 may also include a model training unit 131 and a classification processing unit 132. The model training unit 131 trains the classification model based on the historical manual customer service data as training corpus, and stores the classification model and the corresponding parameter set when the training result meets the requirements. When the classification processing unit 132 receives the text data of the consulting content sent by the consulting content acquisition module 11, it first performs corresponding text processing, such as word segmentation, stop word removal, etc., and then calls the classification model, inputs the processed text into the classification model, and the classification model determines the business type.

[0072] The audit task generation module 14, the audit task allocation module 15 and the reply module 16 serve as audit verification background modules. The audit task pool can be implemented as a task table, for example. When the audit task generation module 14 generates an audit task, it generates a task ID for it, and stores the task ID, the user consultation content and its primary reply content in the task table as an audit task. Of course, the information stored in each audit task can also include other information, such as the client ID, which is used to indicate the client identification of the client that issued the consultation question, etc. The task table can include a task label, and each audit task has a corresponding status and is recorded on the task label. The task table can also be divided into two categories. The first type of task table stores the audit task content to be audited, such as the user consultation content and its primary reply content and the client identification of the client that issued the consultation question, etc. The second type of task table stores the task information that has been audited, which may include specific audit task content, auditor, audit time, etc.

[0073] The audit task assignment module 15 performs task assignment every time an audit task is generated, or scans the task table at regular intervals. When assigning audit tasks to customer service personnel, the priority of assignment can be based on the size of the task processing workload of the customer service end, and the customer service end with the smallest workload can be assigned first, or the order of the customer service end ID can be assigned.

[0074] When the reply module 16 returns the reply content to the user end, one way is to return it in text data format, and the other is to convert the text data into data of the corresponding format according to the received data format. For example, if the consultation content acquisition module 11 receives audio data, the reply module 16 converts the text data into audio data. If the consultation content acquisition module 11 receives video data, the reply module 16 converts the text data into audio data, and then reads a preset consultation video image data, and merges the audio data and the video image data together according to the video parameters to obtain the reply video data. The specific method is as described above, and will not be repeated here.

[0075] The present invention also provides an electronic device, including a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the customer service information processing method of any one of the above embodiments is implemented.

[0076] Figure 5 FIG. 1 is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Figure 5 As shown, the electronic device may include a processor 601 and a memory 602 storing computer program instructions. Figure 5 The electronic device shown may be Figure 1 The first server 1041 in the example may also be the user terminal device 102 or a customer service terminal device.

[0077] Specifically, the processor 601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiment of the present invention.

[0078] The memory 602 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 602 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 602 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 602 is a non-volatile solid-state memory.

[0079] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the methods according to the present invention.

[0080] The processor 601 implements any one of the customer service information processing methods in the above embodiments by reading and executing computer program instructions stored in the memory 602 .

[0081] In one example, the electronic device may further include a communication interface 603 and a bus 610. Figure 5 As shown, the processor 601, the memory 602, and the communication interface 603 are connected and communicate with each other via a bus 610. The electronic device in the embodiment of the present invention may be a server or other computing device, or may be a cloud server.

[0082] The communication interface 603 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiment of the present invention.

[0083] Bus 610 includes hardware, software or both, and the components of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 610 may include one or more buses. Although the embodiment of the present invention describes and shows a specific bus, the present invention considers any suitable bus or interconnection.

[0084] In another aspect, an embodiment of the present invention further provides a computer-readable storage medium having computer program instructions stored thereon; when the computer program instructions are executed by a processor, any one of the customer service information processing methods in the above-mentioned embodiments is implemented, and the computer storage medium is, for example, a memory in the aforementioned electronic device, or a separate device for storing computer program instructions.

[0085] In another aspect, the embodiment of the invention further provides a computer program product, which includes computer program instructions, and when the computer program instructions are executed by a processor, the customer service information processing method described above is implemented. The computer program product is, for example, an application installation package, an application plug-in, a small program, etc.

[0086] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0087] The functional blocks shown in the structural block diagram described above can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier. "Machine-readable medium" can include any medium capable of storing or transmitting information. The machine-readable medium can include a non-transient computer-readable storage medium, such as an electronic circuit, a semiconductor memory device, a ROM, a flash memory, an erasable ROM (EROM), a floppy disk, a CD-ROM, an optical disk, a hard disk, an optical fiber medium, and the machine-readable medium can also include a radio frequency (RF) link, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0088] It should also be noted that the exemplary embodiments mentioned in the present invention describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiments, or in a different order from the embodiments, or several steps can be performed simultaneously.

[0089] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0090] The above is only a specific implementation of the present invention. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system, module and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be covered within the protection scope of the present invention.

Claims

1. A customer service information processing method, characterized in that: Applied to the server, including: When receiving a customer service request from a user, obtaining text data of the user's consultation content from the customer service request; The text data of the user's consultation content is input into the pre-fine-tuned large model and the pre-trained classification model respectively, the large model obtains the primary reply content for the user's consultation content, and the classification model obtains the business type matching the user's consultation content; Generate an audit task based on the user consultation content and its primary reply content, and store the user consultation content and its primary reply content in the audit task into an audit task pool corresponding to the business type; Sending audit tasks in the audit task pool to designated clients of the same business type based on the task allocation strategy; and Receive the review completion reply content returned by the customer service end, and send the review completion reply content to the user end that issued the customer service request.

2. The customer service information processing method according to claim 1, characterized in that: Before sending the reply content of the review completion to the user terminal that issued the customer service request, the method includes: Verify the data format of the user consultation content in the customer service request; and When the data format of the reply content after review is different from the data format of the user consultation content in the customer service request, the data format of the reply content after review is converted into the same data format as the data format of the user consultation content in the customer service request.

3. The customer service information processing method according to claim 2, characterized in that: The data format of the user consultation content is one or more of text format, image format, audio format or video format.

4. The customer service information processing method according to claim 3, characterized in that: When the data format of the user consultation content is in audio format, the text data of the reviewed reply content is converted into audio reply data using audio parameters matching the business type.

5. The customer service information processing method according to claim 3, characterized in that: When the data format of the user consultation content is in video format, determine the image data of the customer service personnel whose image matches the business type, use audio parameters that match the image of the customer service personnel to convert the text data of the reviewed reply content into audio data, and process the image data and audio data according to the preset video parameters to obtain video reply data.

6. The customer service information processing method according to claim 5, characterized in that: When the customer service staff image data whose image matches the business type is determined, a virtual digital person whose image matches the business type is rendered to obtain the image data.

7. The customer service information processing method according to claim 1, characterized in that: Further steps include fine-tuning the large model: Acquire historical manual customer service data, and process the historical manual customer service data to obtain a training data set, wherein each training sample in the training data set is a question-answer pair consisting of a consultation content and a reply content; and Based on the P-tuning fine-tuning technology, the GLM large model is fine-tuned using the training data set to obtain a model parameter set that meets the requirements.

8. The customer service information processing method according to claim 1, characterized in that: Each audit task in the audit task pool has a task tag; when the user consultation content and the primary reply content in the audit task are stored in the audit task pool corresponding to the business type, the audit status of the audited task is set to "unreviewed" in the task tag of the audit task; correspondingly, the task tag of each audit task in the audit task pool is scanned, and the audit tasks with the audit status of "unreviewed" in the task tag are sent to the client end of the same business type; correspondingly, when the reply content of the audit completion is received from the client end, the audit status in the task tag of the corresponding audit task is set to "reviewed".

9. The customer service information processing method according to claim 1, characterized in that: The audit task pool corresponding to the business type includes an unaudited task sub-pool and an audited task sub-pool; when the user consultation content and the primary reply content in the audit task are stored in the audit task pool corresponding to the business type, the audit task is stored in the unaudited task sub-pool; correspondingly, the unaudited task sub-pool is scanned to determine an audit task to be sent to the client end of the same business type; correspondingly, when the reply content of the audit completion returned by the client end is received, the corresponding audit task is transferred from the unaudited task sub-pool to the audited task sub-pool.

10. The customer service information processing method according to claim 1, characterized in that: Before sending the audit task in the audit task pool to the client based on the task allocation strategy, the step of determining the client is further included: Acquire working status information of the client terminal, and determine an available client terminal based on the working status information of the client terminal; According to the priority order of the client terminals, the client terminal with the highest current order among the available client terminals is determined as the designated client terminal; Among them, the priority is sorted in order from low to high according to the amount of pending tasks of the client; or, the priority is sorted in order from low to high according to the client ID.

11. The customer service information processing method according to claim 1, characterized in that: After sending the audit task in the audit task pool to the client, further comprising: Monitoring the processing time of the audit task by the client; In response to the processing time of the audit task by the client end exceeding a threshold, the audit task is sent to the secondary client end, and the processing of the audit task by the client end is terminated.

12. The customer service information processing method according to claim 1, characterized in that: The audit task is sent to the client end of the same business type through a third-party message channel.

13. A customer service information processing device, characterized in that: Applied to the server, including: A consultation content acquisition module is configured to receive a customer service request from a user terminal and obtain text data of the user's consultation content from the customer service request; A large model module is configured to input the text data of the user's consultation content into the pre-fine-tuned large model, and obtain the primary reply content for the user's consultation content from the large model; A classification model module is configured to input the text data of the user's consultation content into a pre-trained classification model, and obtain the business type matching the user's consultation content from the classification model; An audit task generation module is configured to generate an audit task based on user consultation content and primary response content thereof, and store the user consultation content and primary response content in the audit task into an audit task pool corresponding to the business type; An audit task allocation module, configured to send audit tasks in the audit task pool to a client terminal of the same business type based on a task allocation strategy; and The reply module is configured to receive the reply content of the review completion returned by the customer service end, and send the reply content of the review completion to the user end that issued the customer service request.

14. The customer service information processing device according to claim 13, characterized in that: It also includes a customer service module, which is configured to receive the review task and send the reviewed reply content to the reply module after the customer service staff completes the review.

15. A customer service information processing system, characterized in that: include: The user terminal device is configured as a user end, and after receiving the consultation content and instructions input by the user, sends a customer service request to the service end, and includes the user consultation content in the customer service request, and receives the returned reply content; The customer service information processing device as claimed in claim 13, which communicates with the user terminal device through a network as a server, and is configured to obtain the primary reply content corresponding to the user consultation content based on the pre-fine-tuned large model when receiving the customer service request sent by the user terminal device, obtain the business type matching the user consultation content based on the classification model, generate an audit task based on the user consultation content and its primary reply content, and store the user consultation content and its primary reply content in the audit task into the audit task pool corresponding to the business type; Based on the task allocation strategy, the audit tasks in the audit task pool are sent to the client end of the same business type; and sending the reply content reviewed by the customer service end to the user terminal device that issued the customer service request; as well as The customer service terminal device is configured as a customer service terminal, receives the review task and sends the reviewed reply content to the customer service information processing device after the customer service personnel completes the review.

16. An electronic device, characterized in that: It comprises a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the customer service information processing method according to any one of claims 1 to 12 is implemented.

17. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by the processor, the customer service information processing method according to any one of claims 1 to 12 is implemented.

18. A computer program product, characterized in that It includes computer program instructions, which, when executed by a processor, implement the customer service information processing method described in any one of claims 1-12.

Images