A data processing method and device based on knowledge graph

Abstract

The present invention discloses a data processing method and device based on knowledge graph, including: constructing a knowledge graph of a product, generating a dialogue text corresponding to the communication information between the customer service and the user; identifying the characteristic information of the product in the static picture, and judging the status of the product; extracting a series of key frames in the video data, comparing the key frames with the characteristic vector set, and obtaining the product problems pre-expressed by the user using the video data; estimating the probability of failure of the product and the type of failure; obtaining the user's misoperation type for the product based on the difference analysis result, and generating a short video of the correct operation corresponding to the misoperation type based on the knowledge graph. Through the present invention, the user's demands and misoperation types can be grasped, and a short video of the correct operation for the wrong operation can be quickly generated, which greatly improves the timeliness and accuracy of the e-commerce customer service reply.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and specifically to a data processing method and device based on a knowledge graph. Background Art

[0002] In order to improve user satisfaction and enhance competitiveness, modern e-commerce platforms generally set high standards for customer service responses. Many platforms require customer service to respond to user inquiries in a very short time (for example, 30 seconds). Although this immediacy requirement improves user experience, it also puts tremendous pressure on the customer service team. However, in many customer service calls, it is not a problem with the product quality, but a misunderstanding caused by the user's lack of understanding of the product's functions and structure and improper operation. Customer service personnel need to guide users to operate correctly based on the incorrect operations.

[0003] Especially when users send long videos of product usage, customer service personnel not only need to analyze whether there are problems with the product in the video and whether the user's operation of the product is correct, but also need to guide the user to operate and use it correctly, which brings huge response pressure to the customer service team. Therefore, how to accurately analyze user demands and provide product operation guidance in e-commerce customer service has become a technical problem that needs to be solved urgently in this field. Summary of the invention

[0004] In view of the above problems, the present application provides a data processing method based on knowledge graph to solve the technical problem of low response efficiency of e-commerce customer service.

[0005] To achieve the above objectives, the present application provides a data processing method based on knowledge graph, including:

[0006] Construct a knowledge graph of the product, define entities, attributes and relationships in the knowledge graph based on the structure, function and operation of the product, wherein each product is defined as an entity, the structure, function and operation of the product are defined as attributes, and the association between the structure, function and operation of the product is defined as a relationship; store the knowledge graph in a database of a server, and generate a feature vector set of each product in the knowledge graph and cache it in the memory of the server, wherein the feature vector set stores feature vectors of the operation of each product and images of common problems;

[0007] Acquire communication information between customer service and users from an e-commerce customer service platform, the communication information including text information, voice information, static images and video data; use voice recognition technology to convert the voice information into corresponding text, and merge it into the text information in chronological order to obtain a conversation text;

[0008] Using image recognition technology to identify feature information of the product in the static image, combining the feature information with the semantics of the conversation text upstream and downstream of the static image to determine the status of the product; extracting a series of key frames from the video data, and comparing the key frames with the feature vector set to obtain the product problem pre-expressed by the user using the video data;

[0009] Use NLP technology to analyze the semantics of the conversation text and predict the user's problem demands for the product; and estimate the probability of failure of the product and the type of failure based on the problem demands, the product problem and the status of the product; when the probability is greater than a preset value, start a preset product failure handling process; when the probability is less than a preset value, perform a difference analysis on the feature vectors of the image of the product operation in the feature vector set of the corresponding product in the knowledge graph, and obtain the user's misoperation type for the product based on the difference analysis result; access the API interface of the knowledge graph with the misoperation type and the key frame as input parameters, generate a short video of the correct operation corresponding to the misoperation type based on the knowledge graph, and send the short video to the user.

[0010] In order to solve the above technical problems, the present application also provides another technical solution: a data processing device based on knowledge graph, comprising:

[0011] A construction module is used to construct a knowledge graph of products, define entities, attributes and relationships in the knowledge graph based on the structure, function and operation of the products, wherein each product is defined as an entity, the structure, function and operation of the product are defined as attributes, and the association between the structure, function and operation of the product is defined as a relationship; store the knowledge graph in a database of a server, and generate a feature vector set of each product in the knowledge graph and cache it in the memory of the server, wherein the feature vector set stores feature vectors of the operation of each product and images of common problems;

[0012] The extraction module is used to obtain the communication information between the customer service and the user from the e-commerce customer service platform, wherein the communication information includes text information, voice information, static images and video data; the voice information is converted into corresponding text using voice recognition technology, and then merged into the text information in chronological order to obtain the dialogue text;

[0013] An analysis module is used to use image recognition technology to identify feature information of the product in the static image, and to determine the status of the product by combining the feature information and the semantics of the conversation text upstream and downstream of the static image; extract a series of key frames from the video data, and compare the key frames with the feature vector set to obtain the product problem pre-expressed by the user using the video data;

[0014] A processing module is used to use NLP technology to analyze the semantics of the conversation text and predict the user's problem demands for the product; and according to the problem demands, the product problem and the status of the product, estimate the probability of failure of the product and the type of failure; when the probability is greater than a preset value, start a preset product failure processing process; when the probability is less than a preset value, perform a difference analysis on the feature vector of the image of the product operation in the feature vector set of the corresponding product in the knowledge graph, and obtain the user's misoperation type for the product based on the difference analysis result; access the API interface of the knowledge graph with the misoperation type and the key frame as input parameters, generate a short video of the correct operation corresponding to the misoperation type based on the knowledge graph, and send the short video to the user.

[0015] Different from the existing technology, the above technical solution constructs a product knowledge graph based on the data processing method of the knowledge graph. A comprehensive analysis of the communication information between customer service and users can obtain the user's demands for the product, and further combined with the knowledge graph, it can quickly analyze the status of the product and whether it is faulty, and can analyze the type of misoperation of the product in the video data, and finally generate a short video of the correct operation corresponding to the misoperation type. Therefore, in this technical solution, customer service personnel can grasp the user's demands and misoperation types without watching the entire video data, and quickly generate a short video of the correct operation for the wrong operation, which greatly improves the timeliness and accuracy of e-commerce customer service responses.

[0016] The above-mentioned records related to the invention content are only an overview of the technical solution of the present application. In order to enable ordinary technicians in the field to more clearly understand the technical solution of the present application, and then implement it according to the text of the specification and the contents recorded in the drawings, and to make the above-mentioned purpose and other purposes, features and advantages of the present application easier to understand, the following is an explanation in combination with the specific implementation mode and drawings of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings are only used to illustrate the principles, implementation methods, applications, characteristics and effects of the specific embodiments of the present invention and other related contents, and shall not be considered as limiting the present application.

[0018] In the drawings of the specification:

[0019] Figure 1 A flowchart of the data processing method based on the knowledge graph described in the specific implementation method;

[0020] Figure 2 A flowchart of generating a short video based on the knowledge graph in a specific implementation manner;

[0021] Figure 3 It is a flowchart of obtaining the type of incorrect operation based on the difference analysis result in a specific implementation manner;

[0022] Figure 4 A module block diagram of a knowledge graph-based data processing device according to a specific implementation method.

[0023] The reference numerals in the above drawings are described as follows:

[0024] 400. Data processing device based on knowledge graph; 401. Construction module; 402. Extraction module; 403. Analysis module; 404. Processing module. DETAILED DESCRIPTION

[0025] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects that can be achieved, the following is a detailed description of the specific embodiments listed in conjunction with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0026] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the various technical features mentioned in the embodiments can be combined in any way to form a corresponding implementable technical solution.

[0027] Unless otherwise defined, the technical terms used in this document have the same meanings as those generally understood by those skilled in the art to which this application belongs; the use of relevant terms in this document is only for describing specific embodiments and is not intended to limit this application.

[0028] In the description of this application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" logical relationship.

[0029] In the present application, terms such as “first” and “second” are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.

[0030] Without further limitations, in this application, the words "include", "comprises", "has" or other similar open-ended expressions used in sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.

[0031] Similar to the understanding in the Examination Guidelines, in this application, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of this application, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.

[0032] In the description of the embodiments of the present application, space-related expressions used, such as "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "vertical", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicate the orientation or position relationship based on the orientation or position relationship shown in the specific embodiments or drawings, and are only for the convenience of describing the specific embodiments of the present application or facilitating the reader's understanding, and do not indicate or imply that the referred device or component must have a specific position, a specific orientation, or be constructed or operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.

[0033] Unless otherwise expressly specified or limited, in the description of the embodiments of the present application, the terms such as "install", "connect", "connect", "fix", "set", etc. used should be understood in a broad sense. For example, the "connection" can be a fixed connection, a detachable connection, or an integrated setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For technicians in the technical field to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0034] See also Figures 1 to 3, this embodiment provides a data processing method based on knowledge graph. The data processing method based on knowledge graph is applied to e-commerce customer service, and is used to analyze the communication information with users during the customer service process based on the product's knowledge graph, especially to analyze the status of the product and the type of misoperation of the product from the static pictures and video data therein, thereby improving the timeliness and accuracy of the e-commerce customer service response. The data processing method based on knowledge graph can be applied to various domestic e-commerce platforms such as Taobao, JD.com, and Douyin, and can also be applied to various cross-border e-commerce platforms such as Amazon and Meikeduo.

[0035] like Figure 1 As shown, the data processing method based on the knowledge graph in this embodiment includes:

[0036] S101. Construct a knowledge graph of products, define entities, attributes and relationships in the knowledge graph based on the structure, function and operation of the products, wherein each product is defined as an entity, the structure, function and operation of the product are defined as attributes, and the association between the structure, function and operation of the product is defined as a relationship; store the knowledge graph in a database of a server, and generate a feature vector set for each product in the knowledge graph and cache it in the memory of the server, wherein the feature vector set stores feature vectors of the operation of each product and images of common problems.

[0037] S102, obtaining communication information between customer service and users from an e-commerce customer service platform, the communication information including: text information, voice information, static images and video data; using speech recognition technology to convert the voice information into corresponding text, and integrating it into the text information in chronological order to obtain a conversation text.

[0038] S103, judging the status of the product and the product problem pre-expressed by the user using the video data; specifically: using image recognition technology to identify feature information of the product in the static image, combining the feature information and the semantics of the conversation text upstream and downstream of the static image to judge the status of the product; extracting a series of key frames in the video data, and comparing the key frames with the feature vector set to obtain the product problem pre-expressed by the user using the video data.

[0039] S104. Generate a short video of the correct operation corresponding to the type of incorrect operation based on the knowledge graph, and send the short video to the user; specifically: use NLP technology to analyze the semantics of the conversation text to predict the user's problem demands for the product; and estimate the probability of failure of the product and the type of failure based on the problem demands, the product problem and the status of the product; when the probability is greater than a preset value, start a preset product failure handling process; when the probability is less than a preset value, perform a difference analysis on the key frame and the feature vector of the image of the product operation in the feature vector set of the corresponding product in the knowledge graph, and obtain the user's incorrect operation type for the product based on the difference analysis result; access the API interface of the knowledge graph with the incorrect operation type and the key frame as input parameters, generate a short video of the correct operation corresponding to the incorrect operation type based on the knowledge graph, and send the short video to the user.

[0040] In step S101, the product refers to the product sold by the e-commerce merchant, and the knowledge graph can be constructed using a graph database (such as Neo4i) or an RDF-based triple storage. The operations in the product knowledge graph include normal operation and abnormal operation of the product, as well as common problems caused by abnormal operation and their solutions. The common problems of the product and their solutions are used as attributes associated with the product entity. There is a relationship between each common problem and the operation that causes the common problem. For example, for a simple coffee machine operation knowledge graph, the nodes may include steps such as "turn on the machine", "put in coffee powder", "add water", and "start brewing", and each step is associated with a correct operation image.

[0041] When constructing a knowledge graph, relevant data can be collected from multiple channels such as product documents, user feedback, customer service records, etc. Then clean the data to ensure the accuracy and consistency of the data. Based on the structure, function, operation and other properties of the product, define entities, attributes and relationships, and extract entities, attributes and relationships from the collected data. Integrate the extracted knowledge to eliminate contradictions and ambiguities. In this embodiment, a corresponding image is configured for each of the operations and the common problems, that is, operation 1 is configured with operation 1 image, and common problem 1 is configured with problem image 1. The knowledge graph is also provided with a feature vector set for storing feature vectors of each image.

[0042] In step S103, when judging the status of the product, first extract the features of the product from the static image, including the shape, color, position, and prompt words displayed by the product, etc.; and analyze these features; determine the position of the static image in the communication information between the customer service and the user, and expand a certain range (e.g., 5 dialogue texts) upstream and downstream with this position as the center, and then extract keywords related to product quality or problems from the determined range (e.g., keywords include inaccurate measurement, no effect, etc.); finally, filter out features related to the keywords from the features, and compare the filtered features with the knowledge graph to obtain the status of the product. Taking the static image of the ear thermometer measuring temperature as an example, first extract the position and posture of the ear thermometer's detection head (i.e., whether it is facing the measured object) and the information displayed by the ear thermometer (e.g., displayed temperature or prompt words, displayed color) from the static image of the ear thermometer measuring temperature. Then determine the position of the static image of the ear thermometer measuring temperature in the entire communication information, assuming it is in the 20th line, and then analyze whether keywords such as "inaccurate measurement", "high temperature" or "low temperature" appear in lines 16 to 25 of the communication information (assuming that low temperature appears). Finally, filter out features related to the keyword "low temperature" from the extracted features (for example, the ear thermometer is not inserted into the ear canal), and compare the filtered features with the knowledge graph to obtain the status of the product (that is, the ear thermometer functions normally but is improperly operated).

[0043] In step S104, the preset product failure handling process includes a return process and a repair process; based on the product order information and the product warranty information, determine whether the product is within the warranty period, and recommend a return process or the repair process; determine the final process from the processing flow based on the confirmation operation of the customer service staff.

[0044] In step S104, when estimating the probability of the product failing and the type of failure, the types and frequencies of failures can be counted by first analyzing the historical failure data of the product (including the failure type and the corresponding phenomenon), and then the estimates can be made based on the statistics of product failure types in the historical data of the e-commerce customer service and the description of the user's problem with the product. Specifically, a suitable statistical model (such as MIL-HDBK-217F) can be designed to annotate the historical failure data and the historical data of the e-commerce customer service, and then the annotated data can be used to train the statistical model, and finally the problem demands, product problems and product status can be input into the trained statistical model to estimate the probability of failure and the type of failure.

[0045] And in step S104, a RESTful API interface can be designed for the knowledge graph, which allows users to upload pictures of incorrect operations and returns correct operation pictures. The above-mentioned image recognition, knowledge graph query and result generation logic are implemented in the API interface. And the API interface is deployed on the server so that it can be accessed through the network.

[0046] Taking the customer service for coffee machine operation as an example, the communication information between the customer service and the user includes a video of the operation of the coffee machine. In step S103, a series of coffee machine operation frames (i.e., key frames) are extracted. In step S104, it is analyzed that the probability of failure of the coffee machine is lower than a threshold (e.g., 10%). At this time, the extracted coffee machine operation frames are uploaded to the API interface of the knowledge graph.

[0047] The API interface receives the coffee machine operation frame and calls the image recognition model for recognition, identifying the wrong operation step as "no water added". The API interface queries the corresponding correct operation steps and their associated pictures in the knowledge graph based on the identified wrong operation steps. The API interface finds the correct operation step "add water" and its associated pictures in the knowledge graph, and generates a short video of the "add water" operation based on the pictures associated with the step "add water", and generates and returns the short video URL.

[0048] In step S104, the generated short video can be sent to the user through the e-commerce customer service platform, and the user can watch the correct operation method of the product by clicking on the short video.

[0049] like Figure 2 As shown, in the above embodiment, the short video of the correct operation corresponding to the incorrect operation type is generated based on the knowledge graph, including:

[0050] S201, obtaining the input erroneous operation type, and locating a specific erroneous operation node according to the key frame and the properties and relationships of the structure and function of the product in the knowledge graph;

[0051] S202, in the knowledge graph, according to the erroneous operation node, retrieve the correct operation image associated with the erroneous operation node, the correct operation image showing the correct operation steps and status of the product; and analyze the possible causes of the erroneous operation in combination with the product structure and function relationship in the knowledge graph;

[0052] S203. Use video editing software to combine the correct operation image, the operation guidance text, and necessary commentary audio or subtitles to generate a short video of the correct operation.

[0053] like Figure 3As shown, in the above embodiment, the key frame and the feature vector of the image of the product operation in the feature vector set corresponding to the product in the knowledge graph are subjected to difference analysis, and the type of user's misoperation on the product is obtained based on the difference analysis result, including:

[0054] S301, constructing a comparison template for each correct operation of the product based on the feature vector set;

[0055] S302, filtering and sharpening the key frame, and extracting target features that can reflect product operation from the processed key frame using a convolutional neural network model;

[0056] S303, determining each of the target features and a corresponding comparison template, and performing template matching between the target feature and the corresponding comparison template to obtain the difference between the target feature and the corresponding correct operation in terms of operation posture, position, and operation sequence;

[0057] S304: Calculate the data error operation type based on the difference, wherein the error operation type includes any one or more of an operation sequence error, an operation posture error, and a product position error. The convolutional neural network model can use a self-attention mechanism to extract the target feature from the processed key frame.

[0058] like Figure 4 As shown, in one embodiment, a data processing device 400 based on a knowledge graph is provided. The data processing device 400 based on a knowledge graph includes: a construction module 401, an extraction module 402, an analysis module 403 and a processing module 404.

[0059] The construction module 401 is used to construct a knowledge graph of the product, define entities, attributes and relationships in the knowledge graph based on the structure, function and operation of the product, wherein each product is defined as an entity, the structure, function and operation of the product are defined as attributes, and the association between the structure, function and operation of the product is defined as a relationship; the knowledge graph is stored in a database of the server, and a feature vector set of each product in the knowledge graph is generated and cached in the memory of the server, wherein the feature vector set stores feature vectors of the operation of each product and images of common problems.

[0060] The extraction module 402 is used to obtain communication information between customer service and users from the e-commerce customer service platform, and the communication information includes: text information, voice information, static images and video data; use speech recognition technology to convert the voice information into corresponding text, and merge it into the text information in chronological order to obtain the conversation text.

[0061] The analysis module 403 is used to use image recognition technology to identify the feature information of the product in the static image, and to determine the status of the product by combining the feature information and the semantics of the conversation text upstream and downstream of the static image; extract a series of key frames from the video data, and compare the key frames with the feature vector set to obtain the product problems pre-expressed by the user using the video data.

[0062] The processing module 404 is used to use NLP technology to analyze the semantics of the conversation text, predict the user's problem demands for the product; and estimate the probability of failure of the product and the type of failure based on the problem demands, the product problem and the status of the product; when the probability is greater than a preset value, start a preset product failure handling process; when the probability is less than a preset value, perform a difference analysis on the feature vector of the image of the product operation in the feature vector set of the corresponding product in the knowledge graph, and obtain the user's misoperation type for the product based on the difference analysis result; access the API interface of the knowledge graph with the misoperation type and the key frame as input parameters, generate a short video of the correct operation corresponding to the misoperation type based on the knowledge graph, and send the short video to the user.

[0063] In the above embodiment, the generating of a short video of the correct operation corresponding to the incorrect operation type based on the knowledge graph includes:

[0064] Obtaining the input erroneous operation type, and locating a specific erroneous operation node according to the key frame and the properties and relationships of the structure and function of the product in the knowledge graph;

[0065] In the knowledge graph, according to the erroneous operation node, a correct operation image associated with the erroneous operation node is retrieved, wherein the correct operation image displays the correct operation steps and status of the product; and the possible causes of the erroneous operation are analyzed in combination with the product structure and function relationship in the knowledge graph;

[0066] Use video editing software to combine the correct operation images, operation guidance text, and necessary explanatory audio or subtitles to generate a short video of the correct operation.

[0067] In the above embodiment, performing difference analysis between the key frame and the feature vector of the image of the product operation in the feature vector set corresponding to the product in the knowledge graph, and obtaining the type of user's misoperation on the product based on the difference analysis result includes:

[0068] Building a comparison template for each correct operation of the product based on the feature vector set;

[0069] Filtering and sharpening the key frames, and extracting target features that can reflect product operation from the processed key frames using a convolutional neural network model;

[0070] Determine each of the target features and a corresponding comparison template, and perform template matching between the target feature and the corresponding comparison template to obtain the difference between the target feature and the corresponding correct operation in terms of operation posture, position, and operation sequence;

[0071] The data error operation type is calculated based on the difference, and the error operation type includes any one or more of an operation sequence error, an operation posture error, and a product position error.

[0072] In the above embodiment, the convolutional neural network model uses a self-attention mechanism to extract the target features from the processed key frames.

[0073] In the above embodiment, a knowledge graph of the product is constructed, and a comprehensive analysis of the communication information between the customer service and the user can be performed to obtain the user's demands for the product, and further combined with the knowledge graph, the product status and whether it is faulty can be quickly analyzed, and the type of misoperation for the product in the video data can be analyzed, and finally a short video of the correct operation corresponding to the misoperation type can be generated. Therefore, in this technical solution, the customer service staff can grasp the user's demands and misoperation types without watching the entire video data, and quickly generate a short video of the correct operation for the wrong operation, which greatly improves the timeliness and accuracy of the e-commerce customer service response.

[0074] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of this application, this does not limit the scope of patent protection of this application. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of this application using the contents recorded in the specification and drawings of this application, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the scope of patent protection of this application.

Claims

1. A data processing method based on knowledge graph, characterized in that: include: Construct a knowledge graph of the product, define entities, attributes and relationships in the knowledge graph based on the structure, function and operation of the product, wherein each product is defined as an entity, the structure, function and operation of the product are defined as attributes, and the association between the structure, function and operation of the product is defined as a relationship; store the knowledge graph in a database of a server, and generate a feature vector set of each product in the knowledge graph and cache it in the memory of the server, wherein the feature vector set stores feature vectors of the operation of each product and images of common problems; Acquire communication information between customer service and users from an e-commerce customer service platform, the communication information including text information, voice information, static images and video data; use voice recognition technology to convert the voice information into corresponding text, and merge it into the text information in chronological order to obtain a conversation text; Using image recognition technology to identify feature information of the product in the static image, combining the feature information with the semantics of the conversation text upstream and downstream of the static image to determine the status of the product; extracting a series of key frames from the video data, and comparing the key frames with the feature vector set to obtain the product problem pre-expressed by the user using the video data; Use NLP technology to analyze the semantics of the conversation text and predict the user's problem demands for the product; and estimate the probability of failure of the product and the type of failure based on the problem demands, the product problem and the status of the product; when the probability is greater than a preset value, start a preset product failure handling process; when the probability is less than a preset value, perform a difference analysis on the feature vectors of the image of the product operation in the feature vector set of the corresponding product in the knowledge graph, and obtain the user's misoperation type for the product based on the difference analysis result; access the API interface of the knowledge graph with the misoperation type and the key frame as input parameters, generate a short video of the correct operation corresponding to the misoperation type based on the knowledge graph, and send the short video to the user.

2. The data processing method based on knowledge graph according to claim 1 is characterized in that: The generating a short video of the correct operation corresponding to the incorrect operation type based on the knowledge graph includes: Obtaining the input erroneous operation type, and locating a specific erroneous operation node according to the key frame and the properties and relationships of the structure and function of the product in the knowledge graph; In the knowledge graph, according to the erroneous operation node, a correct operation image associated with the erroneous operation node is retrieved, wherein the correct operation image displays the correct operation steps and status of the product; and the possible causes of the erroneous operation are analyzed in combination with the product structure and function relationship in the knowledge graph; Use video editing software to combine the correct operation images, operation guidance text, and necessary explanatory audio or subtitles to generate a short video of the correct operation.

3. The data processing method based on knowledge graph according to claim 2 is characterized in that: The step of performing a difference analysis between the key frame and the feature vector of the image of the product operation in the feature vector set corresponding to the product in the knowledge graph, and obtaining the type of user's misoperation on the product based on the difference analysis result, includes: Building a comparison template for each correct operation of the product based on the feature vector set; Filtering and sharpening the key frames, and extracting target features that can reflect product operation from the processed key frames using a convolutional neural network model; Determine each of the target features and a corresponding comparison template, and perform template matching between the target feature and the corresponding comparison template to obtain the difference between the target feature and the corresponding correct operation in terms of operation posture, position, and operation sequence; The data error operation type is calculated based on the difference, and the error operation type includes any one or more of an operation sequence error, an operation posture error, and a product position error.

4. The data processing method based on knowledge graph according to claim 3 is characterized in that: The convolutional neural network model uses a self-attention mechanism to extract the target features from the processed key frames.

5. The data processing method based on knowledge graph according to claim 1 is characterized in that: The preset product failure handling process includes a return process and a repair process; Determine whether the product is within the warranty period based on the product order information and the product warranty information, and recommend a return process or the repair process; The final process is determined from the processing flow according to the confirmation operation of the customer service personnel.

6. A data processing device based on knowledge graph, characterized in that: include: A construction module is used to construct a knowledge graph of products, define entities, attributes and relationships in the knowledge graph based on the structure, function and operation of the products, wherein each product is defined as an entity, the structure, function and operation of the product are defined as attributes, and the association between the structure, function and operation of the product is defined as a relationship; store the knowledge graph in a database of a server, and generate a feature vector set of each product in the knowledge graph and cache it in the memory of the server, wherein the feature vector set stores feature vectors of the operation of each product and images of common problems; The extraction module is used to obtain the communication information between the customer service and the user from the e-commerce customer service platform, wherein the communication information includes text information, voice information, static images and video data; the voice information is converted into corresponding text using voice recognition technology, and then merged into the text information in chronological order to obtain the dialogue text; An analysis module is used to use image recognition technology to identify feature information of the product in the static image, and to determine the status of the product by combining the feature information and the semantics of the conversation text upstream and downstream of the static image; extract a series of key frames from the video data, and compare the key frames with the feature vector set to obtain the product problem pre-expressed by the user using the video data; A processing module is used to use NLP technology to analyze the semantics of the conversation text and predict the user's problem demands for the product; and according to the problem demands, the product problem and the status of the product, estimate the probability of failure of the product and the type of failure; when the probability is greater than a preset value, start a preset product failure processing process; when the probability is less than a preset value, perform a difference analysis on the feature vector of the image of the product operation in the feature vector set of the corresponding product in the knowledge graph, and obtain the user's misoperation type for the product based on the difference analysis result; access the API interface of the knowledge graph with the misoperation type and the key frame as input parameters, generate a short video of the correct operation corresponding to the misoperation type based on the knowledge graph, and send the short video to the user.

7. The data processing device based on knowledge graph according to claim 6 is characterized in that: The step of generating a short video of the correct operation corresponding to the incorrect operation type based on the knowledge graph includes the following steps: Obtaining the input erroneous operation type, and locating a specific erroneous operation node according to the key frame and the properties and relationships of the structure and function of the product in the knowledge graph; In the knowledge graph, according to the erroneous operation node, a correct operation image associated with the erroneous operation node is retrieved, wherein the correct operation image displays the correct operation steps and status of the product; and the possible causes of the erroneous operation are analyzed in combination with the product structure and function relationship in the knowledge graph; Use video editing software to combine the correct operation images, operation guidance text, and necessary explanatory audio or subtitles to generate a short video of the correct operation.

8. The data processing device based on knowledge graph according to claim 7, characterized in that: The step of performing a difference analysis between the key frame and the feature vector of the image of the product operation in the feature vector set corresponding to the product in the knowledge graph, and obtaining the type of user's misoperation on the product based on the difference analysis result, includes: Building a comparison template for each correct operation of the product based on the feature vector set; Filtering and sharpening the key frames, and extracting target features that can reflect product operation from the processed key frames using a convolutional neural network model; Determine each of the target features and a corresponding comparison template, and perform template matching between the target feature and the corresponding comparison template to obtain the difference between the target feature and the corresponding correct operation in terms of operation posture, position, and operation sequence; The data error operation type is calculated based on the difference, and the error operation type includes any one or more of an operation sequence error, an operation posture error, and a product position error.

9. The data processing device based on knowledge graph according to claim 8, characterized in that: The convolutional neural network model uses a self-attention mechanism to extract the target features from the processed key frames.

Images