After-sales service method and device of medical equipment, electronic equipment and storage medium

Abstract

The application provides a medical equipment after-sales service method and device, electronic equipment and a storage medium. The medical equipment after-sales service information input by a user is acquired, and the after-sales service information is mapped to a vector. The after-sales service type corresponding to the after-sales service information is determined based on a first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes a mapping relationship between a vector corresponding to historical after-sales service information and a historical after-sales service type. The user is provided with after-sales service based on the after-sales service type, the disadvantages of hard classification of a large model are avoided, a fixed classification dimension and range do not need to be set in advance, the accuracy of after-sales service type matching is improved, the service response time is reduced, and the after-sales service efficiency is improved.

Description

Technical Field

[0001] This application belongs to the field of after-sales service technology, and in particular relates to an after-sales service method, apparatus, electronic device and storage medium for medical devices. Background Technology

[0002] In the field of medical equipment after-sales service, with the advancement of digital transformation, users' after-sales service needs are becoming increasingly diversified, and multi-source service data such as equipment repair, video customer service, intelligent customer service, and fault self-diagnosis are constantly accumulating. Existing after-sales service methods for medical equipment mostly rely on large models to classify and match services. However, large models generally suffer from the prominent problem of rigid classification, that is, fixed and unchanging classification dimensions and ranges need to be set in advance, and then user needs are matched through this rigid classification system. It can be seen that the rigid classification of large models lacks flexibility, cannot adapt to the dynamic changes of multi-source service data, and is difficult to cover the diverse and personalized potential service needs of users. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method, apparatus, electronic device and storage medium for after-sales service of medical devices, which can avoid the drawbacks of rigid classification of large models.

[0004] In a first aspect, embodiments of this application provide a method for after-sales service of medical devices, including: Obtain after-sales service information for medical devices input by the user, and map the after-sales service information into a vector; The after-sales service type corresponding to the after-sales service information is determined based on the first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: the mapping relationship between the vector corresponding to the historical after-sales service information and the historical after-sales service type; After-sales service is provided to the user based on the aforementioned after-sales service type.

[0005] In some embodiments, determining the after-sales service type corresponding to the after-sales service information based on the first mapping relationship corresponding to the user and the vector includes: Calculate the similarity between the vector and the vector corresponding to historical after-sales service information; The similarity weight of the historical after-sales service information is determined based on the data type corresponding to the historical after-sales service information. Based on the similarity and similarity weights corresponding to each historical after-sales service information, the target similarity between the vector and the vector corresponding to the historical after-sales service information is obtained; The historical after-sales service type corresponding to the maximum target similarity is determined as the after-sales service type corresponding to the after-sales service information.

[0006] In some embodiments, determining the similarity weight of the historical after-sales service information based on the type corresponding to the historical after-sales service information includes: Obtain the second mapping relationship between the data type and similarity weight of pre-established historical after-sales service information; Based on the second mapping relationship and the data types corresponding to the historical after-sales service information, the similarity weights corresponding to the historical after-sales service information are obtained. The data types include: structured text after-sales service data types, intelligent customer service after-sales service data types, and video after-sales service data types. The second mapping relationship includes: the correspondence between structured text after-sales service data types and the first similarity weight, the correspondence between intelligent customer service after-sales service data types and the second similarity weight, and the correspondence between video after-sales service data types and the third similarity weight. The first similarity weight is greater than the second similarity weight, and the second similarity weight is greater than or equal to the third similarity weight.

[0007] In some embodiments, mapping the after-sales service information into a vector includes: When the after-sales service information is structured text after-sales service data, first target information is extracted from the after-sales service information, and a vector is generated based on the first target information. The first target information includes at least one of the following: user ID, device ID, device model, repair time, and fault information. When the after-sales service information is intelligent customer service after-sales service data, extract the second target information from the after-sales service information and generate a vector based on the second target information. The second target information includes at least one of the following: user ID, user question text, dialogue round, keywords, and the validity of intelligent customer service's question feedback. When the after-sales service information is video after-sales service data, a third target information is extracted from the after-sales service information, and a vector is generated based on the third target information. The third target information includes at least one of the following: user ID, user video dialogue voice, dialogue text, and key repair frame.

[0008] In some embodiments, the method further includes: Obtain the historical after-sales service information and corresponding after-sales service types input by the user, and map the historical after-sales service information into a vector; The vector corresponding to the historical after-sales service information is normalized to obtain the normalized vector. Establish the first mapping relationship between each normalized vector and the corresponding after-sales service type; Store the first mapping relationship.

[0009] In some embodiments, providing after-sales service to the user based on the after-sales service type includes: When the after-sales service type is an inquiry category, the after-sales service information is matched with the knowledge base to obtain the matching result, and an inquiry response is output based on the matching result; When the after-sales service type is a complaint category, after-sales service is provided to the user based on the method corresponding to the complaint category. When the after-sales service type is to switch to manual service, an operation guide is output based on a preset template to access manual service; If the type of after-sales service is not identified, a prompt message is output to guide the user to re-enter the after-sales service information.

[0010] In some embodiments, providing after-sales service to the user based on the complaint classification includes: When the complaint is classified as a quality problem complaint or an after-sales service complaint, the after-sales service information is matched with the knowledge base to obtain a first matching result. The first matching result is then input into the large model to obtain a first response information. The first response information is output, and a first prompt information is also output. The first prompt information is used to indicate that there is a quality problem complaint or an after-sales service complaint. In the case of an emergency complaint, the after-sales service information is matched with the knowledge base to obtain a second matching result. Based on the second matching result, a reassurance dialogue is output, and the target personnel are notified to provide human assistance. If the complaint is classified as "other complaints", the after-sales service information is matched with the knowledge base to obtain a third matching result, and a reassurance dialogue is output based on the third matching result.

[0011] Secondly, embodiments of this application provide an after-sales service device, comprising: The acquisition module is used to acquire after-sales service information for medical devices input by the user and map the after-sales service information into a vector. The determining module is used to determine the after-sales service type corresponding to the after-sales service information based on the first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: the mapping relationship between the vector corresponding to the historical after-sales service information and the historical after-sales service type; The service module is used to provide after-sales service to the user based on the after-sales service type.

[0012] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the methods described above.

[0013] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the methods described above.

[0014] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes an electronic device to execute any of the methods described above.

[0015] This application provides an after-sales service method for medical devices. It acquires after-sales service information for medical devices input by a user and maps this information to a vector. Based on a first mapping relationship corresponding to the user and the vector, it determines the after-sales service type corresponding to the information. The first mapping relationship includes a mapping relationship between vectors corresponding to historical after-sales service information and historical after-sales service types. Providing after-sales service to the user based on the chosen after-sales service type avoids the drawbacks of rigid classification in large models. It eliminates the need to pre-set fixed classification dimensions and ranges, achieving dynamic matching through the first mapping relationship. This adapts to the dynamic changes in multi-source after-sales service data, avoiding service matching biases caused by classification limitations. Based on a user-specific first mapping relationship, the matching results better match the user's actual service needs, improving the accuracy of after-sales service type matching, reducing service response time, and increasing after-sales service efficiency. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram illustrating the implementation process of an after-sales service method for medical devices provided for the purposes of this application; Figure 2 A schematic diagram illustrating the implementation process of step S102 provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the implementation process of an after-sales service method for a medical device provided in this application embodiment; Figure 4 A schematic diagram of the structure of an after-sales service device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0019] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0020] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0021] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once detected," or "in response to detection."

[0022] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0024] Based on the problems in related technologies, this application provides an after-sales service method for medical devices that can be applied to electronic devices, including: mobile phones, tablets, wearable devices, augmented reality (AR) / virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), etc. The electronic devices can act as servers for the after-sales service system. The after-sales service method for medical devices provided in this application can be implemented using the processor of the electronic device.

[0025] Figure 1 A schematic diagram illustrating the implementation process of an after-sales service method for medical devices provided for the purposes of this application is shown below. Figure 1 As shown, the after-sales service methods for medical devices include: Step S101: Obtain after-sales service information for medical devices input by the user, and map the after-sales service information into a vector.

[0026] In this embodiment, after-sales service information refers to various types of information input by users to seek after-sales service for medical devices. It refers to various types of information actively input by users through various terminals (such as mobile apps, computer web pages, offline service terminals, intelligent customer service dialog boxes, etc.) to seek after-sales service support for medical devices. Its form is not limited, including but not limited to text, voice, video, images, structured forms, etc. A vector, also known as a feature vector, is a set of numerical values ​​that can be used for mathematical calculations, similarity comparisons, and association analysis. It is formed by processing unstructured (such as voice, free text, video frames) or structured (such as fixed-format repair orders) after-sales service information through preset feature extraction and encoding algorithms. Its core function is to transform after-sales service information that cannot be directly quantified into a standardized data form that can be recognized and calculated by computers, facilitating subsequent determination of the after-sales service type through similarity analysis, mapping matching, and other operations.

[0027] In this embodiment, after-sales service information for medical devices input by users through various terminals (such as after-sales service portals on mobile apps, repair forms on computer web pages, intelligent customer service dialog boxes, and video customer service call interfaces) can be received. This information can take any form, including text, voice, video, and images. After receiving the information, preliminary preprocessing is performed, including noise reduction (such as removing background noise from voice and invalid characters from text) and format standardization (such as converting voice to text and extracting text from images), to ensure that subsequent vector mapping steps can be executed normally and to avoid invalid information interfering with the processing results. The preprocessed after-sales service information is then mapped into feature vectors.

[0028] In this embodiment, after information preprocessing, a preset encoding algorithm and feature extraction algorithm can be used to quantize the preprocessed after-sales service information and convert it into a standardized feature vector. The encoding algorithm can be a commonly used text encoding algorithm in the field (such as Word2Vec, BERT, TF-IDF, etc.), and the feature extraction algorithm can be a deep learning-based feature extraction algorithm (such as CNN, LSTM, etc.). The specific algorithm selection can be adapted to the type of after-sales service information (such as text, voice, video). For example, for text-based after-sales service information, the BERT algorithm can be used for encoding to extract semantic features from the text and generate a feature vector; for voice-based information, it can be first converted to text and then the above encoding algorithm can be used to generate a vector; for video-based information, voice, text, and keyframe image features can be extracted from the video, fused, and then a vector generated. The final generated feature vector must meet standardization requirements to ensure that the converted vectors of after-sales service information of different forms and dimensions can be uniformly similar and matched.

[0029] Step S102: Determine the after-sales service type corresponding to the after-sales service information based on the first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: the mapping relationship between the vector corresponding to the historical after-sales service information and the historical after-sales service type.

[0030] In this embodiment, the first mapping relationship is a historical data association relationship established for a single user. It is trained and constructed based on the user's past after-sales service behavior data and serves as the core basis for matching current after-sales service needs. Specifically, it refers to the one-to-one correspondence between the feature vector obtained after mapping and transformation of each historical after-sales service information submitted by the user, and the final determined after-sales service type (such as inquiry, complaint, or transfer to human agent) after the actual processing of that historical after-sales service information. This mapping relationship is user-unique, meaning that the first mapping relationship of different users is independent of each other and is constructed only based on their own historical data, ensuring that the matching result matches the needs of a single user. The after-sales service type refers to the category divided according to the nature and content of the user's after-sales service needs, including but not limited to inquiry categories, complaint categories, and transfer to human agent categories, used to clarify the specific execution direction of after-sales service.

[0031] In this embodiment, a user-specific first mapping relationship (pre-stored in a database) can be invoked. This mapping relationship contains vectors corresponding to all historical after-sales service information of the user, and the after-sales service types corresponding to these vectors. Subsequently, the vectors corresponding to the generated after-sales service information are compared and matched with the vectors corresponding to the historical after-sales service information in the first mapping relationship to find the vector corresponding to the historical after-sales service information with the highest similarity to the current vector. Finally, the after-sales service type corresponding to the vector with the highest similarity is determined as the after-sales service type corresponding to the after-sales service information, thereby completing the accurate matching of service types.

[0032] Step S103: Provide after-sales service to the user based on the after-sales service type.

[0033] In this embodiment, after-sales service types may include primary categories and secondary categories under each primary category. Different after-sales service types may have corresponding processing methods.

[0034] In this embodiment, after determining the after-sales service type, a preset processing flow corresponding to that type is invoked to provide targeted after-sales service responses to the user. For example, if the determined service type is an inquiry category, the inquiry service flow is executed, calling the knowledge base for quick matching and outputting a response; if the determined service type is a complaint category, the complaint service flow is executed; if the determined service type is a transfer to human agent category, the transfer to human agent flow is executed, guiding the user to connect with a human customer service representative. After the after-sales service is completed, the after-sales service information, the generated vector, and the determined service type are added to the user's historical data for subsequent updates to the first mapping relationship, achieving closed-loop optimization of the method.

[0035] The method provided in this application embodiment obtains after-sales service information for medical devices input by the user and maps the after-sales service information into a vector; it determines the after-sales service type corresponding to the after-sales service information based on a first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: a mapping relationship between the vector corresponding to historical after-sales service information and historical after-sales service types; providing after-sales service to the user based on the after-sales service type can avoid the drawbacks of rigid classification in large models, without the need to pre-set fixed classification dimensions and ranges, and achieve dynamic matching through the first mapping relationship formed by historical data, adapting to the dynamic changes of multi-source after-sales service data, and avoiding service matching deviations caused by classification limitations; constructing the first mapping relationship based on the user's exclusive historical data makes the matching results more in line with the user's actual service needs, improving the accuracy of after-sales service type matching; and quickly determining the service type through vector mapping and historical association matching, reducing service response time and improving after-sales service efficiency.

[0036] In some embodiments, Figure 2 This is a schematic diagram illustrating the implementation process of step S102 provided in an embodiment of this application, as follows: Figure 2 As shown, step S102 can be achieved through the following steps: Step S1021: Calculate the similarity between the vector and the vector corresponding to the historical after-sales service information.

[0037] In this embodiment, similarity refers to the degree of similarity between the vector corresponding to the current after-sales service information and the feature vector corresponding to each piece of historical after-sales service information in the first mapping relationship. Its core function is to measure the closeness of the association between the current after-sales service demand and the historical after-sales service demand. The higher the similarity value, the more similar the current demand is to the historical demand, and the more suitable the after-sales service type corresponding to the historical demand is as a reference for the service type of the current demand; the lower the similarity value, the looser the association between the two, and the less reference value. Commonly used similarity calculation methods include, but are not limited to, cosine similarity, Euclidean distance, Manhattan distance, etc.

[0038] In this embodiment of the application, a preset similarity calculation algorithm can be used to calculate the similarity between the vector corresponding to the after-sales service information and the vector corresponding to each historical after-sales service information in the first mapping relationship, so as to obtain multiple similarity values ​​consistent with the number of historical vectors.

[0039] Step S1022: Determine the similarity weight of the historical after-sales service information based on the data type corresponding to the historical after-sales service information.

[0040] In this embodiment, the similarity weight refers to a weight coefficient (typically ranging from 0 to 1) assigned to the similarity of each piece of historical after-sales service information based on its data type. Its core function is to distinguish the influence of different data types of historical after-sales service information on the current after-sales service type matching result. A higher weight coefficient indicates greater reference value for the historical after-sales service information, and a higher proportion of its similarity in the overall calculation; conversely, a lower weight coefficient indicates less reference value for the historical after-sales service information, and a lower proportion of its similarity in the overall calculation.

[0041] In this embodiment of the application, after calculating all similarity values, for each piece of historical after-sales service information, the corresponding similarity weight is determined according to the data type of the historical after-sales service information.

[0042] Step S1023: Based on the similarity and similarity weight corresponding to each historical after-sales service information, obtain the target similarity between the vector and the vector corresponding to the historical after-sales service information.

[0043] In this embodiment, the target similarity refers to the comprehensive similarity value obtained by multiplying the similarity and the corresponding similarity weight. It is the final basis for measuring the degree of correlation between the current vector and the vector of the historical after-sales service information. For example, if the similarity of a certain historical after-sales service information is 0.8 and the corresponding similarity weight is 0.6, then the target similarity of the historical after-sales service information is 0.8 × 0.6 = 0.48.

[0044] In this embodiment of the application, for each piece of historical after-sales service information, the calculated similarity value corresponding to the historical after-sales service information is multiplied by the determined similarity weight to obtain the target similarity corresponding to the historical after-sales service information.

[0045] Step S1024: Determine the after-sales service type corresponding to the maximum target similarity as the after-sales service type corresponding to the after-sales service information.

[0046] In this embodiment, after calculating the target similarity values ​​corresponding to all historical after-sales service information, all target similarity values ​​are sorted (preferably from largest to smallest), and the historical after-sales service information with the largest target similarity value is selected. The vector of this historical after-sales service information is the historical data most closely related to the current vector and has the highest reference value. Subsequently, the after-sales service type corresponding to the largest target similarity is directly determined as the after-sales service type corresponding to the current after-sales service information, completing accurate matching of service types. If multiple historical after-sales service information have the same target similarity value and are all the maximum value, the service type corresponding to any one of the historical after-sales service information can be selected, or the after-sales service type can be determined by combining the time proximity of the historical after-sales service information (prioritizing recent historical vectors), further improving the rationality of the matching.

[0047] The method provided in this application achieves accurate vector comparison through similarity calculation, avoiding errors caused by simple matching and improving the accuracy of service type matching. It introduces similarity weights to distinguish the reference value of historical after-sales service information of different data types, solving the problem that historical after-sales service information of different data types has the same impact on the current matching, making the matching results more reasonable. It determines the final after-sales service type by ranking the target similarity, which is logically clear and highly operable, further improving the accuracy and reliability of service type matching and reducing problems such as missing requirements and matching deviations.

[0048] In some embodiments, step S1022 can be implemented by the following steps: Step S221: Obtain the second mapping relationship between the pre-established data type and similarity weight.

[0049] In this embodiment of the application, the second mapping relationship is a pre-established association relationship used to clarify the similarity weight allocation rules.

[0050] Step S222: Based on the second mapping relationship and the data types corresponding to the historical after-sales service information, obtain the similarity weights corresponding to the historical after-sales service information. The data types include: structured text after-sales service data types, intelligent customer service after-sales service data types, and video after-sales service data types. The second mapping relationship includes: the correspondence between structured text after-sales service data types and the first similarity weight, the correspondence between intelligent customer service after-sales service data types and the second similarity weight, and the correspondence between video after-sales service data types and the third similarity weight. The first similarity weight is greater than the second similarity weight, and the second similarity weight is greater than or equal to the third similarity weight.

[0051] In this application embodiment, the structured text after-sales service data type is a specific type of historical after-sales service information. It refers to text-based after-sales service data with a fixed format, fixed fields, strong data standardization, and high information accuracy. Its core characteristics are clear structure and explicit key information, allowing direct extraction of core target information through algorithms without complex feature extraction processes. Specific examples include, but are not limited to, equipment repair forms (containing fixed fields such as user ID, equipment ID, equipment model, repair time, fault description, contact person, and contact information), fault feedback forms, and after-sales service satisfaction evaluation forms. This type of data has high information accuracy and significant reference value. The intelligent customer service after-sales service data type is another specific type of historical after-sales service information. It refers to various after-sales service-related data generated during user interaction with intelligent customer service. This type of data is the intelligent customer service after-sales service data type. Its core characteristics are a certain degree of unstructured nature, variable data format, and a large amount of natural language dialogue content. Intelligent customer service after-sales data includes, but is not limited to, the text of conversations between users and intelligent customer service, the audio of conversations (converted to text), the user's questions, the intelligent customer service's responses, the number of conversation rounds, the conversation duration, and the user's satisfaction feedback on the intelligent customer service's responses. This type of data can reflect the user's real-time needs, but its accuracy and standardization are slightly lower than structured text after-sales service data. Video after-sales service data is the third specific type of historical after-sales service information. It refers to various after-sales service-related data generated when users submit after-sales service requests through video customer service channels. This type of data is characterized by multiple data dimensions, high processing difficulty, and high information extraction complexity, including information in multiple forms such as video, audio, and text. Video after-sales service data includes, but is not limited to, video call recordings between users and video customer service, audio content in the video (converted to text), the text of conversations in the video, keyframe images of the equipment malfunctioning parts shown by the user in the video, and the video duration. This type of data can intuitively reflect the equipment malfunction situation, but information extraction is difficult, standardization is poor, and its reference value is relatively low. First similarity weight, second similarity weight, and third similarity weight: These are the similarity weight values ​​corresponding to the three types of after-sales service data mentioned above, and all three are values ​​between 0 and 1.

[0052] In this embodiment, the structured text after-sales service data type has the strongest standardization, the highest information accuracy, and the greatest reference value, and therefore has the largest corresponding weight; the intelligent customer service after-sales service data type has the second highest standardization and accuracy, and its corresponding weight is in the middle; the video after-sales service data type has the worst standardization, the greatest difficulty in information extraction, and the lowest relative reference value, and therefore has the smallest corresponding weight, ensuring that the weight allocation is positively correlated with the data value and improving the rationality of the matching.

[0053] In this embodiment, corresponding similarity weight values ​​can be set according to the information standardization, accuracy, and reference value of the data types. Specific values ​​can be set based on the relationship that the first similarity weight is greater than the second similarity weight, and the second similarity weight is greater than or equal to the third similarity weight. For example, considering a practical application scenario, the first similarity weight can be set to 0.6, the second similarity weight to 0.3, and the third similarity weight to 0.1. These values ​​can be flexibly adjusted according to the actual data characteristics and application requirements. Finally, a one-to-one correspondence between the three data types and their corresponding weight values ​​is established, forming a second mapping relationship. This mapping relationship is stored in a database for easy subsequent retrieval. An optimization interface is also reserved to periodically update the weight values ​​and optimize the second mapping relationship based on subsequent application effects and data changes.

[0054] In this embodiment, data type identification can be performed on historical after-sales service information. After identifying the data type of the historical after-sales service information, a pre-built second mapping relationship is retrieved from the database. Based on the identified specific data type, a match is made in the second mapping relationship to find the similarity weight value corresponding to that data type. For example, if a piece of historical information is identified as a structured text after-sales service data type, a first similarity weight is obtained; if a piece of historical information is identified as an intelligent customer service after-sales service data type, a second similarity weight is obtained; if a piece of historical information is identified as a video after-sales service data type, a third similarity weight is obtained. The weight value obtained from the match is directly determined as the similarity weight corresponding to that piece of historical after-sales service information and used for the calculation of target similarity. If a piece of historical after-sales service information contains multiple data types (such as both structured text and video), a weighted average method can be used, combined with the corresponding data type weights, to calculate the comprehensive similarity weight of that piece of historical after-sales service information, further improving the rationality of weight allocation.

[0055] The method provided in this application, through clear similarity weight allocation rules, makes weight allocation more reasonable and operable, avoiding subjectivity in weight allocation; the similarity weights cover the main data types, adapt to the needs of multi-source data fusion, and provide support for the efficient utilization of subsequent multi-source data.

[0056] In some embodiments, step S101 can be implemented by the following steps: Step S1011: If the after-sales service information is structured text after-sales service data, extract the first target information from the after-sales service information and generate a vector based on the first target information. The first target information includes at least one of the following: user ID, device ID, device model, repair time, and fault information.

[0057] In this embodiment, the first target information is the most core and critical set of information in the structured text after-sales service data. Its core function is to represent the user's specific after-sales service needs and the basic information of the equipment. It is the core basis for generating the feature vector corresponding to the structured text after-sales service data, ensuring that the generated vector accurately reflects the core content of this type of data. The first target information includes, but is not limited to, at least one of the following: user ID, device ID, device model, repair time, and fault information. The specific functions of each piece of information are as follows: user ID is used to associate user identity, ensuring that the vector uniquely corresponds to the user; device ID is used to locate the specific faulty device and distinguish the after-sales service needs of different devices; device model is used to clarify the specific parameters and specifications of the device, providing a basis for fault judgment and service matching; repair time is used to refer to the user's historical repair cycle to assist in judging the fault type; fault information is used to directly represent the fault phenomenon and fault location of the device, and is the core basis for judging the after-sales service type.

[0058] In this embodiment, a preset text extraction algorithm (such as a regular expression extraction algorithm or a keyword matching extraction algorithm) can be used to extract core information from structured text after-sales service data. Since this type of data has a fixed field format, the corresponding first target information can be quickly extracted directly based on preset field identifiers (such as "User ID:", "Device Model:", "Fault Information:"). For example, from a device repair order, first target information such as "User ID: XXX", "Device ID: XXX", "Device Model: XXX", and "Fault Information: Device cannot be turned on" can be extracted. If the structured text contains multiple first target information, all can be extracted; if only some are contained, the existing target information can be extracted. After extracting the first target information, since some target information is text-based (such as fault information or device model), it cannot be directly used for vector generation. Therefore, the extracted first target information needs to be quantized and encoded. For numerical target information (such as repair time, which can be converted into a timestamp), its numerical form is directly retained. For text-based target information (such as fault information and device model), a preset text encoding algorithm (such as Word2Vec, BERT, etc.) is used to convert it into a standardized numerical code. For identification information (such as user ID and device ID), one-hot encoding or other methods are used to convert it into a numerical form. All the first target information after quantization and encoding is combined in a preset order (such as user ID → device ID → device model → repair time → fault information) to form a set of standardized numerical values, which is the feature vector corresponding to the structured text after-sales service data. For example, the vector generated based on the first target information can be represented as: E_struct = Embedding(user ID, device ID, model, repair time, fault information), such as: [-0.09745866060256958,.......].

[0059] Step S1012: If the after-sales service information is intelligent customer service after-sales service data, extract the second target information from the after-sales service information and generate a vector based on the second target information. The second target information includes at least one of the following: user ID, user question text, dialogue rounds, keywords, and the validity of the intelligent customer service's question feedback.

[0060] In this embodiment, the second target information is the most core and critical set of information in the intelligent customer service after-sales service data. Its core function is to characterize the user's core after-sales service needs, the complexity of those needs, and the reference value of historical dialogues. It is the core basis for generating the corresponding feature vector of the intelligent customer service after-sales service data, ensuring that the generated vector accurately reflects the core content of this type of data. The second target information includes, but is not limited to, at least one of the following: user ID, user question text, dialogue rounds, keywords, and the validity of the intelligent customer service's feedback. The specific functions of each piece of information are as follows: user ID is used to associate user identity; user question text is a direct reflection of the user's core needs and can intuitively reflect the user's after-sales service needs; dialogue rounds are used to reflect the complexity of the user's needs (the more dialogue rounds, the more complex the needs); keywords are used to quickly extract the user's core needs and avoid interference from redundant dialogue content; the validity of the intelligent customer service's feedback is used to judge the reference value of the historical dialogue (valid feedback indicates high reference value, invalid feedback indicates low reference value).

[0061] In this embodiment, natural language processing technology can be used to extract core information from intelligent customer service after-sales service data (mainly dialogue text and speech-to-text content), removing redundant dialogue content (such as irrelevant small talk and repetitive statements). Keyword extraction algorithms (such as TF-IDF algorithm) are used to extract keywords (core needs) from user question text; dialogue turn statistics algorithms are used to count the number of dialogue turns between the user and intelligent customer service; semantic analysis algorithms are used to determine the effectiveness of intelligent customer service's feedback (such as whether the user accepts the intelligent customer service's response); simultaneously, the user ID and complete user question text are extracted and combined to form second target information. The extracted second target information is then quantified to meet the requirements of vector generation: for the user ID, one-hot encoding is used to convert it into numerical form; for the user question text and keywords, text encoding algorithms such as BERT are used to convert them into standardized numerical codes; for the dialogue turns, their numerical form is directly retained (e.g., 3 turns of dialogue equals 3); for the effectiveness of intelligent customer service's feedback, binary encoding is used (1 for effective, 0 for ineffective). All the quantified second target information is combined in a preset order (e.g., user ID → user question text encoding → keyword encoding → dialogue rounds → feedback validity) to form a standardized set of values, which is the feature vector corresponding to this intelligent customer service after-sales service data. For example, the vector generated based on the second target information can be represented as: E_text = Embedding(user ID, user question text, dialogue rounds, keywords, question answer validity), such as: [-0.038637470453977585,.......].

[0062] Step S1013: If the after-sales service information is video after-sales service data, extract the third target information from the after-sales service information and generate a vector based on the third target information. The third target information includes at least one of the following: user ID, user video dialogue voice, dialogue text, and key repair frame.

[0063] In this embodiment, the third target information is the most core and critical set of information in the video after-sales service data. Its core function is to characterize the user's core after-sales service needs and the equipment's fault status. It is the core basis for generating the corresponding feature vector of the video after-sales service data, ensuring that the generated vector accurately reflects the core content of this type of data. The third target information includes, but is not limited to, at least one of the following: user ID, user video dialogue voice, dialogue text, and key repair frames. The specific functions of each piece of information are as follows: the user ID is used to associate the user's identity; the user video dialogue voice (after being converted to text) and dialogue text can intuitively reflect the user's core after-sales service needs and the equipment fault phenomena; key repair frames refer to key scenes in the video that can clearly show the faulty parts and fault phenomena of the equipment (such as the scene where the user shows the faulty parts of the equipment), and their corresponding feature information can intuitively reflect the fault status of the equipment and assist in judging after-sales service needs.

[0064] In this embodiment, a video decomposition algorithm can be used to decompose the complete video after-sales service data (video file) into three independent parts: audio, text, and keyframe images. This facilitates the extraction of core information separately: the audio content in the video is converted into text using a speech-to-text algorithm (such as ASR algorithm); the dialogue text in the video is extracted (if there are subtitles, they are extracted directly; otherwise, they are obtained through speech-to-text conversion); and key frame extraction algorithms (such as frame difference method) are used to extract key repair frames in the video that clearly show the faulty parts and phenomena of the equipment (such as the image of the user showing the faulty parts of the equipment), thereby obtaining the third target information. Extracting the third target information from the decomposed audio text, dialogue text, and keyframe images may include: extracting the user ID (which can be extracted from the identifier in the dialogue text and video image); extracting the audio-to-text and video dialogue text as the core representation of the user's needs; and using image feature extraction algorithms (such as CNN algorithm) to extract the core feature information (such as the outline of the faulty parts, color changes, etc.) in the key repair frame images as a representation of the equipment fault situation, and combining them to form the third target information. The extracted third target information is encoded in multiple dimensions to ensure that all information can be converted into numerical form. One-hot encoding can be used for the user ID; BERT algorithm can be used for the voice text and dialogue text; and CNN algorithm can be used for the image feature information of key repair frames, converting them into standardized numerical codes. Subsequently, feature fusion algorithms (such as concatenation fusion and weighted fusion) are used to fuse the numerical information encoded in different dimensions into a unified set of numerical values, thereby generating a vector. The fused numerical set is then standardized (such as Min-Max normalization) to adjust the numerical range, forming a set of standardized feature vectors, which are the feature vectors corresponding to this video after-sales service data. For example, the vector generated based on the third target information can be represented as: E_video = Embedding(User ID, User Video Dialogue Voice, Dialogue Text, Key Repair Frame), such as: [0.009550556540489197,.......].

[0065] The method provided in this application adopts differentiated vector generation methods for different after-sales service information, avoiding the problem that a single generation method cannot adapt to multi-source data, and ensuring that the generated vectors can accurately represent the core content of various types of information; it clarifies the target extraction range of various types of information, improves the targeting and efficiency of feature extraction, and reduces the interference of useless information; it realizes the quantitative conversion of multi-source after-sales service data, and provides a unified data foundation for subsequent similarity calculation and mapping matching.

[0066] In some embodiments, a hierarchical embedding architecture can be used to generate vectors. This architecture can include: a first layer: a raw symbol embedding layer that maps discrete symbols (such as words, IDs, timestamps, image keyframes, repair request text, etc.) into initial dense vectors. A second layer: a context-aware embedding layer that fuses local contextual information, making the embeddings context-sensitive (e.g., distinguishing "right arm" from referring to a robotic arm). A third layer: a cross-modal aligned embedding layer that projects embeddings from different modalities onto a shared semantic space, achieving image-text consistency and structure-text alignment. A fourth layer: a task-oriented aggregation embedding layer that fuses multimodal aligned vectors to generate a final representation for the "fault prediction" task. This hierarchical embedding architecture can be used to generate vectors for all three types of after-sales service information.

[0067] In some embodiments, prior to step S101, the method further includes: Step S1: Obtain the historical after-sales service information and corresponding after-sales service type input by the user, and map the historical after-sales service information into a vector.

[0068] In this embodiment, historical after-sales service data of users can be collected. The after-sales service type is the type ultimately determined based on the processing results (such as inquiry, quality complaint, transfer to human agent, etc.). After completing the historical data collection and preprocessing, a vector generation method is used to extract the corresponding target information (first, second, and third target information) for each specific type of historical after-sales service information (structured text, intelligent customer service, video). Then, through corresponding encoding and fusion methods, each piece of historical after-sales service information is mapped and converted into a corresponding feature vector.

[0069] Step S2: Normalize the vector corresponding to the historical after-sales service information to obtain the normalized vector.

[0070] In this embodiment of the application, normalization processing can uniformly map the values ​​of all vectors to a fixed range of 0-1, eliminating the interference caused by numerical differences and making similarity calculation more accurate.

[0071] In this embodiment, the numerical range and dimensionality of the feature vectors generated from different types of historical after-sales service information may vary (e.g., the numerical range of structured text vectors is 0-100, and the numerical range of video vectors is 0-1000). If these are directly used to construct the first mapping relationship, it will affect the accuracy of subsequent similarity calculations. Therefore, a preset normalization algorithm is used to normalize all historical feature vectors. Commonly used normalization algorithms include Min-Max normalization and Z-Score normalization.

[0072] Step S3: Establish the first mapping relationship between each normalized vector and the corresponding after-sales service type.

[0073] In this embodiment of the application, after obtaining the standardized vectors corresponding to all historical after-sales service information, each normalized standardized vector is associated with the "confirmed after-sales service type" corresponding to the historical after-sales service information to construct a first mapping relationship exclusive to the user.

[0074] Step S4: Store the first mapping relationship.

[0075] In this embodiment, the established user-specific first mapping relationship can be stored in a preset database, using the user ID as an index for easy subsequent retrieval (i.e., during subsequent matching, the user's first mapping relationship can be quickly retrieved based on the user ID). Encrypted storage is employed to ensure the security of user data. Since the normalized vector in the first mapping relationship also needs to be normalized after the after-sales service information is mapped to a vector, this normalization process is also required.

[0076] The method provided in this application normalizes historical vectors, eliminating interference from numerical differences between different vectors and improving the accuracy of subsequent similarity calculations and service type matching. It also enables the storage and updating of the first mapping relationship, and can dynamically optimize the mapping relationship based on the user's historical service data, making the matching results more in line with the user's latest needs.

[0077] In some embodiments, step S103 can be implemented by the following steps: Step S1031: If the after-sales service type is an inquiry category, match the after-sales service information with the knowledge base to obtain a matching result, and output an inquiry response based on the matching result.

[0078] In this embodiment, the inquiry category refers to the user's after-sales service request being a consultation-type request. The inquiry category can be a primary category, and can also include secondary categories. For example, secondary categories may include: equipment usage training, troubleshooting, error codes, common questions, and other inquiries. Inquiry categorization does not require complex service intervention. The knowledge base stores various after-sales service-related knowledge content, including equipment user manuals, troubleshooting guides, and after-sales service procedures, used to quickly match inquiry-type requests and provide answers.

[0079] In this embodiment, a preset knowledge base can be invoked to match current after-sales service information (such as user inquiry text) with the knowledge base. A large model (such as an LLM) is then invoked to organize the matching results into standardized inquiry-response pairs, which are then output to the user to fulfill the response to the inquiry-based request. After the inquiry is completed, the question can be stored in the database, and user feedback information can be collected. The large model can then be fine-tuned based on the feedback information.

[0080] Step S1032: If the after-sales service type is a complaint category, provide after-sales service to the user based on the method corresponding to the complaint category.

[0081] In this embodiment of the application, the complaint classification refers to the user's after-sales service needs being complaint-related needs. The complaint classification can be a primary classification, and can also include secondary classifications, such as: quality problems, after-sales service, emergency complaints, etc.

[0082] In this embodiment of the application, the complaint handling mechanism is activated according to the preset processing flow corresponding to the complaint category.

[0083] Step S1033: If the after-sales service type is to switch to manual service, output operation instructions based on a preset template to enter manual service.

[0084] In this embodiment, "transfer to human assistance" refers to situations where a user's after-sales service needs are complex and cannot be met by intelligent matching, or where the user explicitly requests human assistance and needs to be transferred to a human customer service representative. The preset template is a pre-set operation guide text used to guide users to human assistance, clearly informing users of the steps to transfer to human assistance, waiting time, and other information.

[0085] In this embodiment, a preset operation guide template can be invoked to output the template content (such as "Your needs are quite complex, you have been transferred to a human customer service representative, please wait patiently, the estimated waiting time is XX minutes") to the user. At the same time, the human customer service transfer process is automatically triggered to synchronize user information and after-sales service information to the human customer service representative, thereby improving the efficiency of human service.

[0086] Step S1034: If the after-sales service type is not identified, input prompt information, which is used to guide the user to re-enter the after-sales service information.

[0087] In this embodiment, the prompt information is used to guide the user to re-enter the text of the after-sales service information. When the current service type cannot be identified, the user is prompted to supplement the information and standardize the input to ensure that the service type can be accurately matched in the future.

[0088] In this embodiment of the application, if a clear after-sales service type is not identified (e.g., the user input information is vague or cannot be matched with historical data in the first mapping relationship): a preset prompt message is generated (e.g., "The after-sales service information you entered is unclear. Please supplement the details of the equipment failure or specific needs so that we can provide you with accurate service"), and output to the user to guide the user to re-enter clear and standardized after-sales service information to ensure that the service type can be accurately matched in the future.

[0089] The method provided in this application provides differentiated after-sales service operations for different types of service needs, improving the targeting and efficiency of after-sales service and avoiding poor user experience caused by a "one-size-fits-all" service model. It clarifies the specific processing paths for various service types, making the after-sales service process more standardized, more operable, and easier to implement. A guidance mechanism for unidentified types is set up to reduce missing needs, improve user experience, and ensure accurate matching of service types subsequently. Inquiry-type needs can be quickly responded to through a knowledge base, reducing the pressure on human customer service. Transferring to human categorization enables information synchronization, improving the efficiency of human service, thus balancing service efficiency and user experience.

[0090] In some embodiments, step S1032 can be implemented by the following steps: Step S321: If the complaint is classified as a quality problem complaint or an after-sales service complaint, the after-sales service information is matched with the knowledge base to obtain a first matching result. The first matching result is then input into the large model to obtain a first response information. The first response information is output, and a first prompt information is output. The first prompt information is used to indicate that there is a quality problem complaint or an after-sales service complaint.

[0091] In this embodiment, a quality problem complaint refers to a complaint initiated by a user due to defects or malfunctions in the equipment itself (not caused by improper use). An after-sales service complaint refers to a complaint initiated by a user due to issues such as attitude, efficiency, or handling results during the after-sales service process. The first notification message is a notification to relevant staff (such as a complaint handling specialist) to promptly address and handle quality problem or after-sales service complaints. This first notification message can be output via email, chat software, etc.

[0092] In this embodiment, the current user's after-sales service information (complaint content, user information, equipment information, etc.) is matched with the service knowledge base to obtain a first matching result. The first matching result is input into the large model, and the large model generates a first response information (including reassurance content, processing commitment, estimated processing time, etc.) that fits the complaint type. The first response information is output to the user, and a first prompt information is also output and sent to the relevant complaint handling specialist to remind them to handle the complaint in a timely manner and track the processing progress.

[0093] In this embodiment of the application, a prompt message can be output via API to indicate creditworthiness.

[0094] Step S322: If the complaint is classified as an emergency complaint, the after-sales service information is matched with the knowledge base to obtain a second matching result. Based on the second matching result, a reassurance dialogue is output, and the target personnel are notified to provide manual service.

[0095] In this embodiment, "emergency complaint" refers to a complaint that is urgent and requires immediate handling, such as equipment malfunction causing significant losses or posing a threat to personal safety. "Target personnel" refers to relevant staff responsible for handling emergency complaints, such as customer service representatives or complaint handling supervisors, to ensure a rapid response to emergency complaints.

[0096] In this embodiment, a second matching result can be obtained through a knowledge base. Based on the second matching result, a large model can output a reassuring dialogue to quickly soothe the user's emotions. At the same time, the target personnel (emergency complaint handling specialist) can be notified through system notifications (such as SMS or internal messages) to synchronize the user complaint information and related details to the target personnel, guiding them to immediately contact the user and provide manual service, ensuring that emergency complaints are responded to quickly and handled in a timely manner.

[0097] Step S323: If the complaint is classified as other complaints, the after-sales service information is matched with the knowledge base to obtain a third matching result, and a reassurance dialogue is output based on the third matching result.

[0098] In this embodiment, "other complaints" refers to types of complaints other than quality issues, after-sales service, and emergency complaints, such as complaints about service prices or service scope. "Reassurance dialogue" refers to standardized dialogue text used to soothe the user's emotions, such as "We are very sorry for the bad experience. We will handle your complaint as soon as possible and give you a satisfactory answer."

[0099] In this embodiment of the application, the user's complaint content is matched in the knowledge base to obtain a third matching result. After obtaining the third matching result, a reassuring dialogue can be output through a large model.

[0100] The method provided in this application further subdivides complaint categories and performs differentiated processing for different types of complaints, improving the targeting and rationality of complaint handling and avoiding inefficiency and user dissatisfaction caused by using the same processing procedure for all complaints. Urgent complaints are prioritized to ensure rapid response to urgent issues, reduce user losses, and improve user satisfaction. Response information is generated using a large model, balancing the standardization and efficiency of responses, while soothing dialogue alleviates user emotions and improves user experience. The method clarifies the responsibilities and procedures for handling various types of complaints, ensuring a closed-loop complaint handling process, avoiding issues such as missed complaints and untimely processing, further optimizing after-sales service quality, and enhancing user loyalty.

[0101] Based on the foregoing embodiments, this application provides a specific example, including: Obtain user A's historical after-sales service information for the past year, including 3 structured text after-sales service data entries (equipment repair orders), 2 intelligent customer service after-sales service data entries (dialogue records), and 1 video after-sales service data entry (video customer service dialogue), as well as the after-sales service type corresponding to each historical information entry (2 inquiries, 3 transfers to human agents, and 1 quality complaint); extract the target information of each type of information and generate corresponding vectors; use the Min-Max normalization algorithm to normalize all historical vectors; establish a first mapping relationship between the normalized vectors and the corresponding service types and store it in the database; simultaneously, establish a second mapping relationship, setting the first similarity weight to 0.6, the second similarity weight to 0.3, and the third similarity weight to 0.1.

[0102] User A inputs after-sales service information: "My device (ID: XXX, Model: XXX) cannot be turned on, and I would like to understand the cause of the malfunction." This information is structured text after-sales service data. The first target information (User ID, Device ID, Device Model, Fault Information) is extracted, and a corresponding vector is generated. The cosine similarity between the current vector and all historical normalized vectors in the first mapping relationship is calculated. Based on the second mapping relationship, a corresponding similarity weight is assigned to each historical vector, and the target similarity is calculated. The historical vector corresponding to the maximum target similarity is selected, and its corresponding service type is "Inquiry Classification." Therefore, the current service type is determined to be Inquiry Classification. The current inquiry content is matched with the knowledge base to obtain the matching result of "Common reasons and troubleshooting methods for device malfunction," which is then compiled into an inquiry response and output to User A, completing the after-sales service. If user A inputs after-sales service information as "the equipment malfunctioned after one month of use, and the after-sales service was delayed for half a month, causing serious dissatisfaction", then the service type is determined to be "complaint classification - after-sales service complaint". The complaint content is input into the large model, the first response information (appeasement + handling commitment) is generated and output to user A, and the first prompt information is output to notify the complaint handling specialist to handle it.

[0103] This embodiment achieves accurate matching and efficient response of after-sales service through the above process, avoids the drawbacks of rigid classification of large models, and improves the quality of after-sales service and user experience.

[0104] This application provides another specific example, including: The current input text (e.g., "right arm operation delay, image desynchronization") is processed using the same embedding process as historical data to generate a query vector, and L2 normalization is performed. The cosine similarity between the input text vector and all repair request vectors in the historical database is calculated. The top-1 or top-K historical vectors with the highest similarity are selected. If the maximum similarity exceeds a threshold (e.g., 0.85), a highly similar historical case is considered to exist. The original repair record corresponding to this historical case is then deduced from the case, and a structured prediction result, such as the after-sales service type, is output.

[0105] The method provided in this application does not rely on a predefined fault classification system, but is dynamically generated based on the semantic similarity of real historical behavior, and has strong interpretability and clinical operability.

[0106] Figure 3 A schematic diagram illustrating the implementation process of an after-sales service method for medical devices provided in this application embodiment is shown below. Figure 3 As shown, user input is matched to determine the after-sales service type. Each type has a corresponding processing method. For inquiry categories such as error codes, user training, troubleshooting, common questions, or other inquiries, the system retrieves information from the knowledge base and uses the LLM (Local Management Model) to output responses. For complaint categories such as quality issues and after-sales service, the system notifies users via API, retrieves information from the knowledge base, and uses the LLM to output responses. For urgent complaints, notifications are sent via API / SMS, and the system retrieves information from the knowledge base to output reassuring dialogue, notify human intervention, adjust the priority of human intervention, and finally provide human assistance. For other complaints, the system retrieves information from the knowledge base, uses the LLM to generate reassuring dialogue, and outputs reassuring information. For transferring to human assistance, a template is used to guide the user through the process, leading to human assistance. For other categories, a dialogue guides the user to re-enter information. Data from processed urgent complaints can be analyzed to create user profiles. Data analysis can be performed on data from the after-sales service methods for medical equipment, including: problem entry into the database, collection of user feedback, archiving, annotation, fine-tuning, etc., before finally storing the data in the database.

[0107] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0108] According to the foregoing embodiments, this application provides an after-sales service device. The various modules and units included in the device can be implemented by a processor in a computer device; of course, they can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor unit (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0109] This application provides an after-sales service device. Figure 4 This is a schematic diagram of the structure of an after-sales service device provided in an embodiment of this application, as shown below. Figure 4 As shown, the after-sales service device 400 includes: The acquisition module 401 is used to acquire after-sales service information for medical devices input by the user and map the after-sales service information into a vector. The determining module 402 is used to determine the after-sales service type corresponding to the after-sales service information based on the first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: the mapping relationship between the vector corresponding to the historical after-sales service information and the historical after-sales service type; Service module 403 is used to provide after-sales service to the user based on the after-sales service type.

[0110] In some embodiments, the determining module includes: The first calculation unit is used to calculate the similarity between the vector and the vector corresponding to the historical after-sales service information. The first determining unit is used to determine the similarity weight corresponding to the historical after-sales service information based on the data type corresponding to the historical after-sales service information. The second calculation unit is used to obtain the target similarity between the vector and the vector corresponding to the historical after-sales service information based on the similarity and similarity weight corresponding to each historical after-sales service information. The second determining unit is used to determine the historical after-sales service type corresponding to the maximum target similarity as the after-sales service type corresponding to the after-sales service information.

[0111] In some embodiments, the first determining unit includes: Obtain sub-units to obtain a second mapping relationship between pre-established data types and similarity weights; A subunit is obtained to obtain the similarity weight corresponding to the historical after-sales service information based on the second mapping relationship and the data type corresponding to the historical after-sales service information. The data types include: structured text after-sales service data type, intelligent customer service after-sales service data type, and video after-sales service data type. The second mapping relationship includes: the correspondence between structured text after-sales service data type and the first similarity weight, the correspondence between intelligent customer service after-sales service data type and the second similarity weight, and the correspondence between video after-sales service data type and the third similarity weight. The first similarity weight is greater than the second similarity weight, and the second similarity weight is greater than or equal to the third similarity weight.

[0112] In some embodiments, the acquisition module includes: The first mapping unit is used to extract first target information from the after-sales service information when the after-sales service information is structured text after-sales service data, and generate a vector based on the first target information, wherein the first target information includes at least one of: user ID, device ID, device model, repair time and fault information. The second mapping unit is used to extract second target information from the after-sales service information when the after-sales service information is intelligent customer service after-sales service data, and generate a vector based on the second target information. The second target information includes at least one of: user ID, user question text, dialogue round, keywords, and the validity of intelligent customer service question feedback. The third mapping unit is used to extract third target information from the after-sales service information when the after-sales service information is video after-sales service data, and generate a vector based on the third target information, wherein the third target information includes at least one of: user ID, user video dialogue voice, dialogue text and key repair frame.

[0113] In some embodiments, the after-sales service device further includes: The information acquisition module is used to acquire the historical after-sales service information and corresponding after-sales service type input by the user, and map the historical after-sales service information into a vector; The normalization processing module is used to normalize the vector corresponding to the historical after-sales service information to obtain the normalized vector. A module is established to create the first mapping relationship between each normalized vector and the corresponding after-sales service type. A storage module is used to store the first mapping relationship.

[0114] In some embodiments, the service module includes: The first service unit is used to match the after-sales service information with the knowledge base when the after-sales service type is an inquiry category, obtain the matching result, and output an inquiry response based on the matching result. The second service unit is used to provide after-sales service to the user based on the method corresponding to the complaint category when the after-sales service type is a complaint category. The third service unit is used to output operation instructions based on a preset template when the after-sales service type is classified as manual service, so as to access manual service. The fourth service unit is used to output a prompt message when the after-sales service type is not identified. The prompt message is used to guide the user to re-enter the after-sales service information.

[0115] In some embodiments, the second service unit includes: The first service subunit is used to match the after-sales service information with the knowledge base when the complaint is classified as a quality problem complaint or an after-sales service complaint, to obtain a first matching result, input the first matching result into the large model to obtain a first response information, output the first response information, and output a first prompt information, the first prompt information being used to indicate that there is a quality problem complaint or an after-sales service complaint. The second service subunit is used to match the after-sales service information with the knowledge base when the complaint is classified as an emergency complaint, obtain a second matching result, output a reassurance dialogue based on the second matching result, and notify the target personnel to provide manual service. The third service subunit is used to match the after-sales service information with the knowledge base when the complaint is classified as other complaints, to obtain a third matching result, and to output a reassurance dialogue based on the third matching result.

[0116] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0117] In addition, the after-sales service device and post-operative prediction device mentioned above can be a software unit, a hardware unit, or a combination of software and hardware. They can also be integrated into electronic devices as independent accessories, or exist as independent terminal devices.

[0118] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0119] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device of this embodiment may include: at least one processor 30 ( Figure 5 Only one processor 30, memory 31, and computer program 32 stored in memory 31 and executable on at least one processor 30 are shown. When the processor 30 executes the computer program 32, it implements the steps in any of the above method embodiments, or the processor 30 executes the computer program 32 to implement the functions of each module / unit in the above device or system embodiments.

[0120] For example, computer program 32 may be divided into one or more modules / units, one or more of which are stored in memory 31 and executed by processor 30 to complete this application. One or more modules / units may be a series of computer program 32 instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in an electronic device.

[0121] This application provides an after-sales service system, including: the electronic device and the ultrasonic transceiver module in the above embodiments, wherein the electronic device is communicatively connected to the ultrasonic transceiver module.

[0122] This application also provides a computer-readable storage medium storing a computer program 32, which, when executed by a processor 30, implements the steps described in the above-described method embodiments.

[0123] This application provides a computer program product that, when run on an electronic device, enables the electronic device to perform the steps described in the various method embodiments above.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program 32 instructing related hardware. The computer program 32 can be stored in a computer-readable storage medium, and when executed by the processor 30, it can implement the steps of the various method embodiments described above. The computer program 32 includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a terminal, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0125] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0126] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0127] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0128] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0129] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for after-sales service of medical equipment, characterized in that, include: Obtain after-sales service information for medical devices input by the user, and map the after-sales service information into a vector; The after-sales service type corresponding to the after-sales service information is determined based on the first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: the mapping relationship between the vector corresponding to the historical after-sales service information and the historical after-sales service type; After-sales service is provided to the user based on the aforementioned after-sales service type.

2. The method according to claim 1, characterized in that, The step of determining the after-sales service type corresponding to the after-sales service information based on the first mapping relationship corresponding to the user and the vector includes: Calculate the similarity between the vector and the vector corresponding to historical after-sales service information; The similarity weight of the historical after-sales service information is determined based on the data type corresponding to the historical after-sales service information. Based on the similarity and similarity weights corresponding to each historical after-sales service information, the target similarity between the vector and the vector corresponding to the historical after-sales service information is obtained; The historical after-sales service type corresponding to the maximum target similarity is determined as the after-sales service type corresponding to the after-sales service information.

3. The method according to claim 2, characterized in that, The step of determining the similarity weight corresponding to the historical after-sales service information based on the data type corresponding to the historical after-sales service information includes: Obtain the second mapping relationship between the pre-established data types and similarity weights; Based on the second mapping relationship and the data types corresponding to the historical after-sales service information, the similarity weights corresponding to the historical after-sales service information are obtained. The data types include: structured text after-sales service data types, intelligent customer service after-sales service data types, and video after-sales service data types. The second mapping relationship includes: the correspondence between structured text after-sales service data types and the first similarity weight, the correspondence between intelligent customer service after-sales service data types and the second similarity weight, and the correspondence between video after-sales service data types and the third similarity weight. The first similarity weight is greater than the second similarity weight, and the second similarity weight is greater than or equal to the third similarity weight.

4. The method according to claim 1, characterized in that, The step of mapping the after-sales service information into a vector includes: When the after-sales service information is structured text after-sales service data, first target information is extracted from the after-sales service information, and a vector is generated based on the first target information. The first target information includes at least one of the following: user ID, device ID, device model, repair time, and fault information. When the after-sales service information is intelligent customer service after-sales service data, extract the second target information from the after-sales service information and generate a vector based on the second target information. The second target information includes at least one of the following: user ID, user question text, dialogue rounds, keywords, and the validity of the intelligent customer service's question feedback. When the after-sales service information is video after-sales service data, a third target information is extracted from the after-sales service information, and a vector is generated based on the third target information. The third target information includes at least one of the following: user ID, user video dialogue voice, dialogue text, and key repair frame.

5. The method according to claim 1, characterized in that, The method further includes: Obtain the historical after-sales service information and corresponding after-sales service types input by the user, and map the historical after-sales service information into a vector; The vector corresponding to the historical after-sales service information is normalized to obtain the normalized vector. Establish the first mapping relationship between each normalized vector and the corresponding after-sales service type; Store the first mapping relationship.

6. The method according to any one of claims 1 to 5, characterized in that, Providing after-sales service to the user based on the aforementioned after-sales service type includes: When the after-sales service type is an inquiry category, the after-sales service information is matched with the knowledge base to obtain the matching result, and an inquiry response is output based on the matching result; When the after-sales service type is a complaint category, after-sales service is provided to the user based on the method corresponding to the complaint category. When the after-sales service type is to switch to manual service, an operation guide is output based on a preset template to access manual service; If the type of after-sales service is not identified, a prompt message is output to guide the user to re-enter the after-sales service information.

7. The method according to claim 6, characterized in that, The method of providing after-sales service to the user based on complaint classification includes: When the complaint is classified as a quality problem complaint or an after-sales service complaint, the after-sales service information is matched with the knowledge base to obtain a first matching result. The first matching result is then input into the large model to obtain a first response information. The first response information is output, and a first prompt information is also output. The first prompt information is used to indicate that there is a quality problem complaint or an after-sales service complaint. In the case of an emergency complaint, the after-sales service information is matched with the knowledge base to obtain a second matching result. Based on the second matching result, a reassurance dialogue is output, and the target personnel are notified to provide human assistance. If the complaint is classified as "other complaints", the after-sales service information is matched with the knowledge base to obtain a third matching result, and a reassurance dialogue is output based on the third matching result.

8. An after-sales service device, characterized in that, include: The acquisition module is used to acquire after-sales service information for medical devices input by the user and map the after-sales service information into a vector. The determining module is used to determine the after-sales service type corresponding to the after-sales service information based on the first mapping relationship corresponding to the user and the vector, wherein the first mapping relationship includes: the mapping relationship between the vector corresponding to the historical after-sales service information and the historical after-sales service type; The service module is used to provide after-sales service to the user based on the after-sales service type.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images