E-commerce platform order data intelligent analysis system and method

Abstract

The invention relates to the field of order analysis, and particularly discloses an intelligent analysis system and method for order data of an e-commerce platform, and the method comprises the steps: firstly obtaining evaluation data and order sales data of an e-commerce platform order, then carrying out the feature extraction and correlation analysis of the evaluation data and the order sales data through a deep learning technology, and finally carrying out the analysis through a generator. And a user satisfaction analysis report is generated, so that the customer experience is improved, and an enterprise is helped to improve the product and service quality.

Description

Technical Field

[0001] The present application relates to the field of order analysis, and more specifically, to an e-commerce platform order data intelligent analysis system and method. Background Art

[0002] Online e-commerce refers to online shopping for consumers, transactions between merchants, and online payments that can be carried out through browsers or server platforms in an Internet environment without face-to-face communication between buyers and sellers. With the popularity of online shopping and the growth of demand, various e-commerce platforms have ushered in huge development opportunities. However, this has also triggered fierce competition among platforms. In this competitive environment, companies need to continuously improve the quality of products and services by analyzing order data.

[0003] Among them, understanding the satisfaction of order customers is crucial to improving e-commerce products and services. Traditional satisfaction analysis methods usually rely on manual surveys and basic data statistics, which often fail to fully and accurately reflect all aspects of customer experience.

[0004] Therefore, an intelligent analysis system and method for order data of an e-commerce platform is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an e-commerce platform order data intelligent analysis system and method, which first obtains the evaluation data and order sales data of the e-commerce platform orders, and then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally generates a user satisfaction analysis report through a generator, thereby improving customer experience, thereby helping enterprises improve product and service quality.

[0006] According to one aspect of the present application, there is provided an e-commerce platform order data intelligent analysis system, which includes: The e-commerce platform order data acquisition module is used to obtain the evaluation data and order sales data of the e-commerce platform orders; An e-commerce platform order data extraction module, used to extract optimized order evaluation multimodal semantic feature vectors and order sales text semantic feature vectors from the evaluation data of the e-commerce platform orders and the order sales data; A user satisfaction analysis report generating module is used to generate a user satisfaction analysis report based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector.

[0007] According to another aspect of the present application, a method for intelligent analysis of order data on an e-commerce platform is provided, comprising: Obtain evaluation data and order sales data of e-commerce platform orders; Extracting optimized order evaluation multimodal semantic feature vectors and order sales text semantic feature vectors from the evaluation data of the e-commerce platform orders and the order sales data; A user satisfaction analysis report is generated based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector.

[0008] Compared with the existing technology, the present application provides an e-commerce platform order data intelligent analysis system and method, which first obtains the evaluation data and order sales data of the e-commerce platform orders, and then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally generates a user satisfaction analysis report through a generator, thereby improving customer experience and helping enterprises improve the quality of products and services. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a block diagram of an e-commerce platform order data intelligent analysis system according to an embodiment of the present application.

[0011] Figure 2 Schematic diagram of a block diagram of an e-commerce platform order data extraction module in an e-commerce platform order data intelligent analysis system according to an embodiment of the present application.

[0012] Figure 3 Schematic diagram of a block diagram of an evaluation feature extraction unit at the picture-sharing end in an e-commerce platform order data intelligent analysis system according to an embodiment of the present application.

[0013] Figure 4 This is a block diagram of a user satisfaction analysis report generation module in an e-commerce platform order data intelligent analysis system according to an embodiment of the present application.

[0014] Figure 5 Flow chart of the method for intelligent analysis of e-commerce platform order data according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0016] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0017] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0018] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0019] Figure 1 This is a block diagram of the e-commerce platform order data intelligent analysis system according to the embodiment of the present application. Figure 1 As shown, according to the e-commerce platform order data intelligent analysis system 100 of the embodiment of the present application, it includes: an e-commerce platform order data acquisition module 110, which is used to obtain the evaluation data and order sales data of the e-commerce platform order; an e-commerce platform order data extraction module 120, which is used to extract the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector from the evaluation data of the e-commerce platform order and the order sales data; a user satisfaction analysis report generation module 130, which is used to generate a user satisfaction analysis report based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector.

[0020] In the above-mentioned e-commerce platform order data intelligent analysis system 100, the e-commerce platform order data acquisition module 110 is used to obtain the evaluation data and order sales data of the e-commerce platform order. It should be understood that online e-commerce refers to the activities such as consumers shopping, merchants trading and online payment on the Internet through a browser or server platform without face-to-face. With the popularity of online shopping and the growth of demand, various e-commerce platforms have ushered in huge development opportunities, but they are also facing fierce competition. In this competitive environment, companies need to continuously improve the quality of products and services by analyzing order data. In particular, understanding the satisfaction of order customers is crucial to improving e-commerce products and services. Traditional satisfaction analysis methods usually rely on manual research and basic data statistics, and these methods are often difficult to fully and accurately reflect all aspects of customer experience. Therefore, in the technical solution of this application, the evaluation data and order sales data of e-commerce platform orders are obtained, and combined with deep learning technology, the customer experience is optimized by generating a user satisfaction analysis report, thereby helping companies improve the quality of products and services.

[0021] Specifically, it is crucial for companies to obtain evaluation data and order sales data of e-commerce platform orders. Evaluation data reflects consumers' actual experience and attitude towards products and services, and is a direct source of user feedback. By analyzing evaluation data, companies can understand the advantages and disadvantages of products, user needs and expectations, and thus adjust product strategies, improve service quality, and enhance user satisfaction and loyalty in a timely manner. On the other hand, order sales data provides detailed transaction information and sales trends, including important indicators such as sales, product category preferences, and geographical distribution. By comprehensively analyzing evaluation data and sales data, companies can formulate comprehensive marketing strategies, accurately predict market demand, make data-driven decisions, and enhance market competitiveness and long-term sustainable development capabilities.

[0022] In the above-mentioned e-commerce platform order data intelligent analysis system 100, the e-commerce platform order data extraction module 120 is used to extract optimized order evaluation multimodal semantic feature vectors and order sales text semantic feature vectors from the evaluation data of the e-commerce platform order and the order sales data. It should be understood that the semantic feature vector analysis of comprehensive evaluation data and sales data not only helps companies to deeply understand consumer behavior and market demand, but also provides a scientific basis for product innovation and market promotion. Through these precise data mining and analysis, companies can achieve more accurate market positioning, more effective product strategy adjustments, improve market competitiveness and customer satisfaction, and promote sustainable development and long-term success of enterprises.

[0023] Figure 2 FIG. 1 is a block diagram of an e-commerce platform order data extraction module in an e-commerce platform order data intelligent analysis system according to an embodiment of the present application. Figure 2As shown, in the above-mentioned e-commerce platform order data intelligent analysis system 100, in a specific embodiment of the present application, the e-commerce platform order data extraction module 120 includes: a customer service evaluation feature extraction unit 121, which is used to perform customer service evaluation feature extraction on the evaluation data of the e-commerce platform order to obtain a customer service evaluation semantic feature vector; a picture-sharing end evaluation feature extraction unit 122, which is used to perform picture-sharing end evaluation feature extraction on the evaluation data of the e-commerce platform order to obtain a picture-sharing end evaluation two-dimensional semantic feature vector; an order evaluation multimodal construction unit 123, which is used to construct the customer service evaluation semantic feature vector and the picture-sharing end evaluation two-dimensional semantic feature vector to obtain the optimized order evaluation multimodal semantic feature vector; an order sales text feature extraction unit 124, which is used to perform feature extraction on the order sales data to obtain the order sales text semantic feature vector.

[0024] It should be understood that the semantic feature vector of customer service evaluation reflects the semantic content and emotional color of the interaction between consumers and customer service. By analyzing these feature vectors, enterprises can identify consumers' needs and satisfaction with services, thereby improving customer service processes and improving the efficiency and quality of customer service teams. Among them, the process of extracting customer service evaluation features usually includes extracting key words and expressions of customer service quality from evaluation data, and using natural language processing technology to perform text analysis and sentiment analysis. These feature vectors are not just simple text conversions, but abstract expressions processed by deep learning models or text mining algorithms. They can capture the complex information contained in the evaluation, such as attitudes, emotional tendencies, question types, etc., thereby providing enterprises with targeted improvement suggestions and strategic directions. Through the analysis of semantic feature vectors of customer service evaluation, enterprises can quickly identify and respond to customer needs, improve product design and service processes, and enhance customer satisfaction and loyalty.

[0025] Furthermore, the evaluation data of e-commerce platform orders is subjected to the feature extraction of the evaluation data of the photo-sharing end, mainly to fully understand and analyze the feedback of consumers on the visual evaluation of the product. The two-dimensional semantic feature vector of the evaluation data of the photo-sharing end can extract key information and semantic features from the evaluation, which usually includes consumers' direct feelings and evaluations on the product appearance, function, usage scenarios, etc. Through these data, enterprises can deeply understand consumers' actual usage experience and feelings about product design.

[0026] Furthermore, the text evaluation on the customer service side provides a detailed description of the user experience, while the image evaluation on the photo sharing side provides direct visual evidence. By fusing these two feature vectors, a comprehensive, multi-dimensional evaluation vector can be obtained. This vector not only integrates the emotional information in the text, but also combines the actual feedback information in the image, so as to more comprehensively and accurately reflect the user's real evaluation.

[0027] In particular, feature extraction of order sales data to obtain semantic feature vectors of order sales text is to deeply analyze and understand the key text information of order sales, so as to provide effective support for the market decision-making and strategy optimization of enterprises. Among them, order sales data usually contains information such as product descriptions, sales strategies, promotional activities, customer needs, etc., which are crucial for enterprises to understand market trends, identify sales patterns and optimize product strategies. Through feature extraction, enterprises can convert text information in sales data into structured semantic feature vectors, which can capture the core semantics, keywords and implicit information in the text.

[0028] In a specific embodiment of the present application, the customer service evaluation feature extraction unit 121 includes: extracting customer service evaluation data from the evaluation data of the e-commerce platform order; segmenting the customer service evaluation data to obtain a customer service evaluation text sequence; passing the customer service evaluation text sequence through a customer service evaluation bidirectional long short-term memory neural network model to obtain the customer service evaluation semantic feature vector.

[0029] It should be understandable that customer service evaluation data helps companies achieve personalized customer care. By analyzing consumer evaluations, we can understand their preferences, needs and purchasing motivations, thereby providing customers with a more personalized service experience. This personalized service not only enhances customer satisfaction, but also helps to improve customer loyalty and repeat purchase rate, which has a positive role in promoting the long-term development of the company.

[0030] Furthermore, word segmentation is a key technology in natural language processing. By segmenting continuous text data into meaningful vocabulary units, text data can be made easier for computer processing and analysis. Word segmentation can convert customer service evaluation text into a structured data form, that is, a sequence of words. This allows companies to use existing natural language processing technologies and tools, such as word frequency statistics, keyword extraction, sentiment analysis, etc., to conduct in-depth mining and analysis of customer service evaluations. For example, word frequency statistics can help companies understand the most frequently used keywords and phrases in customer evaluations, thereby identifying the most common types of questions or needs. Word segmentation can help identify and understand semantic and emotional information in text. Different word and phrase combinations can reflect consumers' attitudes, emotional tendencies, and overall evaluations of service experience. Through sentiment analysis technology, companies can quantify customer satisfaction levels, identify positive reviews and negative feedback, and provide specific directions and suggestions for improving service quality and enhancing customer experience.

[0031] Furthermore, the customer-side evaluation text sequence is processed through the customer-side evaluation bidirectional long short-term memory neural network (Bi-LSTM) model to obtain the customer-side evaluation semantic feature vector, in order to accurately capture the contextual information and semantic relationship in the text, thereby providing a more accurate consumer feedback analysis. The bidirectional long short-term memory neural network (Bi-LSTM) model can effectively process the contextual information of text data. Unlike traditional unidirectional neural networks, the Bi-LSTM model can capture the complex relationship and contextual dependency between words in the text by considering both the forward and backward information of the text. This bidirectional processing capability enables the model to fully understand the semantic content in consumer evaluations, avoiding the contextual information that may be ignored by the unidirectional model, thereby improving the accuracy and expressiveness of the semantic feature vector.

[0032] Figure 3 Schematic diagram of a block diagram of a feature extraction unit for evaluating the image sharing end in an intelligent analysis system for order data of an e-commerce platform according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the picture-sharing end evaluation feature extraction unit 122 includes: a picture-sharing end evaluation data extraction subunit 1221, which is used to extract the picture-sharing end evaluation data from the evaluation data of the e-commerce platform order; a picture-sharing end evaluation semantic encoding subunit 1222, which is used to semantically encode the picture-sharing end evaluation data to obtain the picture-sharing end evaluation two-dimensional text feature matrix; a picture-sharing end evaluation two-dimensional convolutional encoding subunit 1223, which is used to pass the picture-sharing end evaluation two-dimensional text feature matrix through a picture-sharing end evaluation two-dimensional text word granularity feature extractor to obtain the picture-sharing end evaluation two-dimensional semantic feature vector.

[0033] It should be understandable that the evaluation data on the photo sharing end provides a visual display of the actual usage scenarios of the product. Through the product evaluations uploaded by consumers, companies can intuitively understand the performance of the product in different environments, including information on appearance design, quality details, and usage effects. This visual feedback can help companies accurately grasp consumers' usage habits and product expectations, and provide a practical basis for product design and improvement.

[0034] Furthermore, semantic coding of the evaluation data on the photo sharing side can effectively utilize modern natural language processing and computer technology to extract useful semantic information from a large amount of content, providing enterprises with deeper market insights and decision support. Semantic coding can convert the evaluation data on the photo sharing side into a text representation with numerical semantic features. This text feature matrix can capture the key semantic information in each photo sharing evaluation, such as descriptive words, emotional expressions, and product feature mentions. These features can be extracted and encoded through modern natural language processing technologies (such as word embedding, BERT, etc.), thereby achieving a deep understanding and analysis of the content of the photo sharing evaluation.

[0035] Furthermore, the two-dimensional text feature matrix of the evaluation on the photo-sharing end is converted into a two-dimensional semantic feature vector of the evaluation on the photo-sharing end through the two-dimensional text word granularity feature extractor of the evaluation on the photo-sharing end, in order to further refine and abstract the semantic information of the visual evaluation data so that it can be more effectively used for deep data analysis and model application. The two-dimensional text word granularity feature extractor of the evaluation on the photo-sharing end can extract representative and information-rich word granularity features from the text feature matrix. These features are not only the representation of a single word or phrase, but also include the semantic relationship and contextual information between words. Through an effective feature extractor, the most critical and representative words for the semantic expression in the photo-sharing end evaluation can be screened out, thereby reducing redundant information and improving the compactness and expression ability of the semantic feature vector. Specifically, each layer of the word granularity feature extractor for evaluating the two-dimensional text on the picture-printing end is used to perform convolution processing, mean pooling processing based on the local feature matrix, and nonlinear activation processing on the input data in the forward transfer of the layer, so that the last layer of the word granularity feature extractor for evaluating the two-dimensional text on the picture-printing end outputs the two-dimensional semantic feature vector for evaluation on the picture-printing end, wherein the input of the word granularity feature extractor for evaluating the two-dimensional text on the picture-printing end is the two-dimensional text feature matrix for evaluation on the picture-printing end.

[0036] In a specific embodiment of the present application, the picture-printing end evaluation semantic encoding subunit 1222 includes: passing the extracted picture-printing end evaluation data through a picture-printing end evaluation text context encoder to obtain a plurality of picture-printing end evaluation feature vectors; and arranging the plurality of picture-printing end evaluation feature vectors in two dimensions to obtain the picture-printing end evaluation two-dimensional text feature matrix.

[0037] It should be understood that the text context encoder can capture and encode the context information in the picture-sharing end evaluation, so as to more accurately understand the intentions and emotions of consumers. The picture-sharing end evaluation not only contains text descriptions, but also potential information in images and contexts. After being processed by the context encoder, this information can obtain more refined feature vectors. These vectors can better reflect the real views and experiences of consumers on the product, and thus provide more accurate market feedback for enterprises. Using the text context encoder to generate multiple feature vectors is helpful to refine and hierarchically analyze the picture-sharing end evaluation data. Each feature vector can represent different aspects of the picture-sharing end evaluation, such as emotional tendency, product function description, usage scenario, etc. Through the analysis of these feature vectors, enterprises can more comprehensively grasp the needs and preferences of consumers, thereby providing accurate data support for product optimization, market strategy adjustment, etc. Specifically, the embedding layer of the picture-sharing end evaluation text context encoder is used to convert the extracted picture-sharing end evaluation data into an embedding vector to obtain a sequence of picture-sharing end embedding vectors; the converter-based Bert model of the picture-sharing end evaluation text context encoder is used to perform global context semantic encoding on the sequence of the picture-sharing end embedding vectors to obtain multiple picture-sharing end evaluation feature vectors.

[0038] Furthermore, arranging multiple picture-sharing end evaluation feature vectors in two dimensions can systematically present the feature information of the evaluation data. When processing picture-sharing evaluations, each evaluation usually contains features of multiple dimensions, such as sentiment, product function description, and usage scenarios. Through two-dimensional arrangement, these features can be organized in rows and columns to form a structured matrix. This matrix format not only clearly shows the multi-dimensional features of each evaluation, but also facilitates the comparison and analysis of different evaluations.

[0039] In a specific embodiment of the present application, the order evaluation multimodal construction unit 1223 includes: associating the customer service end evaluation semantic feature vector and the picture sharing end evaluation two-dimensional semantic feature vector to obtain an order evaluation multimodal semantic feature vector; based on the order sales text semantic feature vector, performing a fine-grained distribution difference-based converter-like structure optimization on the order evaluation multimodal semantic feature vector to obtain an optimized order evaluation multimodal semantic feature vector.

[0040] It should be understood that the purpose of associating the semantic feature vector of customer service evaluation and the two-dimensional semantic feature vector of picture sharing evaluation to obtain the multimodal semantic feature vector of order evaluation is to comprehensively analyze the multi-faceted evaluation and feedback of consumers on the product, so as to fully understand the user's feelings, needs and satisfaction. Among them, the semantic feature vector of customer service evaluation reflects the emotions, attitudes and evaluation of service quality expressed by consumers in text communication. These data help enterprises understand the effect of customer interaction with customer service and direct feedback on service experience. The two-dimensional semantic feature vector of picture sharing evaluation reflects consumers' views and feelings on product appearance, function and other aspects through visual elements, providing intuitive visual feedback. By associating the semantic feature vector of customer service evaluation and the two-dimensional semantic feature vector of picture sharing evaluation, the weight and correlation degree of text evaluation and visual evaluation can be comprehensively considered through hybrid models or data fusion technology, so as to construct a multimodal semantic feature vector of order evaluation. This comprehensive analysis can not only deepen the understanding of consumers' perception of products, but also provide enterprises with more comprehensive and accurate market feedback and product improvement suggestions.

[0041] In particular, in the technical solution of the present application, the order evaluation data includes customer service evaluation and picture sharing evaluation, which are mainly processed into semantic feature vectors and two-dimensional semantic feature vectors through text feedback. The customer service evaluation uses a bidirectional long short-term memory neural network to extract the semantic features of the text sequence, while the picture sharing evaluation is processed by a text context encoder and a word granularity feature extractor, and the obtained two-dimensional semantic feature vector focuses on the spatial and contextual features in the text. Relatively speaking, the order sales data processing focuses on the word embedding and convolutional neural network of the sales text. The generated text semantic feature vector mainly focuses on the time series and content characteristics of the sales data, and more reflects the statistical and trend information of the sales. It has strong quantitative and structural characteristics, but lacks deep semantic information of details and context. This difference in feature processing leads to inconsistent distribution of the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text in the feature space. In the process of fusing these two types of features to obtain the order satisfaction feature vector, due to the difference in the distribution of these two types of features on the feature manifold, the fused order satisfaction feature vector may have abnormal stacking on the boundary of the feature manifold. Therefore, in the technical solution of the present application, based on the semantic feature vector of the order sales text, the order evaluation multimodal semantic feature vector is subjected to a transformer-like structure optimization based on fine-grained distribution differences to obtain an optimized order evaluation multimodal semantic feature vector.

[0042] Among them, based on the semantic feature vector of the order sales text, the order evaluation multimodal semantic feature vector is optimized by a simulated converter structure based on fine-grained distribution difference to obtain an optimized order evaluation multimodal semantic feature vector, including: calculating the position difference between the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector to obtain an e-commerce platform order shift position difference representation vector; after cascading the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector into an e-commerce platform order joint feature vector, the e-commerce platform order joint feature vector is input into a one-dimensional convolutional layer to obtain an e-commerce platform order shift convolutional coding representation vector; after using the first e-commerce platform order linear embedding weight matrix to perform e-commerce platform order linear embedding modulation coding on the order evaluation multimodal semantic feature vector to obtain an order evaluation multimodal semantic feature vector linear modulation vector, the order evaluation multimodal semantic feature vector is calculated. The positional difference between the linear modulation vector of the single evaluation multimodal semantic feature vector and the semantic feature vector of the order sales text is obtained to obtain the shift modulation differential representation vector of the e-commerce platform order; the shifted positional differential representation vector of the e-commerce platform order is used as the value vector, the shifted convolutional coding representation vector of the e-commerce platform order and the shift modulation differential representation vector of the e-commerce platform order are used as the query vector and the key vector, and the shifted positional differential representation vector of the e-commerce platform order, the shifted convolutional coding representation vector and the shift modulation differential representation vector of the e-commerce platform order are input into the shift corrector based on the simulated converter structure to obtain the multi-scale hierarchical fusion shift modulation weight vector of the e-commerce platform order; the positional point multiplication between the multi-scale hierarchical fusion shift modulation weight vector of the e-commerce platform order and the order evaluation multimodal semantic feature vector is calculated to obtain the optimized order evaluation multimodal semantic feature vector.

[0043] The following optimization formula is used to represent that the order evaluation multimodal semantic feature vector is optimized based on the fine-grained distribution difference based on the order sales text semantic feature vector to obtain the optimized order evaluation multimodal semantic feature vector; Wherein, the optimization formula is: ; ; ; ; ;in, Indicates that the order shift of the e-commerce platform is represented by the position difference vector, represents the multimodal semantic feature vector of order evaluation, Represents the semantic feature vector of order sales text, represents the shift convolutional coding representation vector of e-commerce platform orders, represents the shift modulation difference representation vector of the e-commerce platform order, represents the linear embedding weight matrix of the first e-commerce platform order, represents the e-commerce platform order modulation bias vector, represents the multi-scale hierarchical fusion shift modulation weight vector of e-commerce platform orders, represents the transposed vector of the shift-modulated differential representation vector of the e-commerce platform order, Represents the length of the shift convolutional coding representation vector of e-commerce platform orders represents the difference by position, represents a one-dimensional convolutional layer, represents a cascade of the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector, represents the linear rectification function, represents matrix multiplication, represents the normalized exponential function, It means point multiplication by position. Represents the optimized multimodal semantic feature vector of order reviews.

[0044] In the technical solution of the present application, in the process of fusing the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector, the fine-grained inconsistency of feature distribution between the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector is taken into consideration, which will cause abnormal stacking of the fused order satisfaction feature vector on the boundary of the feature manifold. Based on this, in the technical solution of the present application, before fusing the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector to obtain the order satisfaction feature vector, the order evaluation multimodal semantic feature vector is optimized based on the fine-grained distribution difference simulated converter structure based on the order sales text semantic feature vector.

[0045] In the technical solution of the present application, a converter-like structure optimization method based on fine-grained distribution difference is proposed to address the problem of fine-grained inconsistency in feature distribution that occurs in the process of fusing order evaluation multimodal semantic feature vectors and order sales text semantic feature vectors. The method aims to reduce the abnormal stacking phenomenon of order satisfaction feature vectors on the boundary of the feature manifold, so as to improve the robustness and classification performance of the model.

[0046] Specifically, the method first calculates the element-by-element difference between the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text to obtain the element-by-element differential representation vector of the e-commerce platform order shift, thereby capturing the difference between the two feature vectors. Then, the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text are cascaded into an e-commerce platform order joint feature vector and input into a one-dimensional convolutional layer to generate an e-commerce platform order shift convolutional coding representation vector. That is, the deep-level shift implicit information between the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text is captured by one-dimensional convolutional coding. At the same time, the order evaluation multimodal semantic feature vector is subjected to e-commerce platform order linear embedding modulation coding using the first e-commerce platform order linear embedding weight matrix to be a linear modulation vector of the multimodal semantic feature vector of the order evaluation, and the element-by-element difference between the linear modulation vector of the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text is calculated to obtain the shift modulation differential representation vector of the e-commerce platform order. The e-commerce platform order shift modulation differential representation vector is used to represent the position expression of the shift information of the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector in the linear space. That is, in the technical solution of the present application, the e-commerce platform order shift position difference, the e-commerce platform order shift convolutional coding representation vector and the e-commerce platform order shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector in different feature spaces.

[0047] Next, the shifted element-by-element differential representation vector of e-commerce platform orders is used as the value vector, the shifted convolutional coding representation vector of e-commerce platform orders and the shifted modulation differential representation vector of e-commerce platform orders are used as the query vector and the key vector respectively, and these vectors are input into the shift corrector based on the simulated transformer structure to generate the multi-scale hierarchical fusion shift modulation weight vector of e-commerce platform orders. In this way, the importance of features of different scales is evaluated by the attention mechanism, that is, the attention mechanism is used to simulate the hierarchical precipitation expression of the fine-grained difference feature distribution between the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text in different feature spaces. Finally, the element-by-element product of the multi-scale hierarchical fusion shift modulation weight vector of e-commerce platform orders and the multimodal semantic feature vector of order evaluation is calculated to obtain the optimized multimodal semantic feature vector of order evaluation, that is, the weighted operation is used to assign weights to each element in the multimodal semantic feature vector of order evaluation, which reduces the influence of local disturbances and improves the quality of the feature vector.

[0048] In summary, in the method, the position difference of the e-commerce platform order shift, the convolutional coding representation vector of the e-commerce platform order shift, and the modulation difference representation vector of the e-commerce platform order shift respectively represent the explicit expression of the fine-grained distribution difference shift information between the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text in different feature spaces. The generation of the multi-scale hierarchical fusion shift modulation weight vector of the e-commerce platform order uses the attention mechanism to evaluate the importance of features at different scales and provide weight information for feature shift correction. Ultimately, the generation of the optimized multimodal semantic feature vector of order evaluation reduces the impact of disturbances, improves the robustness of the model to local disturbances in the input data, enhances the model's adaptability to noisy or variant data sets, improves the fusion quality of the feature vector, and improves the overall classification performance and training efficiency of the model.

[0049] In a specific embodiment of the present application, the order sales text feature extraction unit 1224 includes: passing the order sales data through an order sales data word embedding encoder to obtain an order sales text word vector sequence; passing the order sales text word vector sequence through an order sales convolutional neural network based on a multi-scale neighborhood feature extraction module to obtain the order sales text semantic feature vector.

[0050] It should be understood that in a typical sales process, order information is often time-series, that is, orders at different time points reflect different market and customer behaviors. Through word vector sequences, we can capture the temporal dynamic changes and trends in order data, helping companies predict seasonal sales fluctuations, changes in product popularity, and changing trends in market demand, thereby optimizing inventory management and production scheduling.

[0051] Furthermore, the word vector sequence of order sales text is converted into semantic feature vector of order sales text through order sales convolutional neural network based on multi-scale neighborhood feature extraction module, in order to further deepen the understanding and utilization of semantic information of sales data. This process combines the advantages of convolutional neural network (CNN) in text processing, and can effectively extract hierarchical features in text data, thereby improving the expression ability and analysis accuracy of data. The design based on multi-scale neighborhood feature extraction module enables convolutional neural network to capture local and global features of order sales text at different scales. Text information in sales data often has different levels of importance and semantic combinations, such as product description, sales conditions, customer feedback, etc. Through multi-scale feature extraction, the network can pay attention to long and short text features at the same time, effectively capture the complex semantic relationship and important information in order data, and make the generated semantic feature vector of order sales text richer and more accurate. Specifically, the order sales convolutional neural network based on the multi-scale neighborhood feature extraction module includes: a first convolutional layer, a second convolutional layer parallel to the first convolutional layer, and a cascade layer connected to the first convolutional layer and the second convolutional layer, wherein the first convolutional layer uses a one-dimensional convolutional kernel with a first scale, and the second convolutional layer uses a one-dimensional convolutional kernel with a second scale. More specifically, the first convolutional layer of the order sales convolutional neural network based on the multi-scale neighborhood feature extraction module is used to perform one-dimensional convolution encoding on the order sales text word vector sequence to obtain a first-scale feature vector; the second convolutional layer of the order sales convolutional neural network based on the multi-scale neighborhood feature extraction module is used to perform one-dimensional convolution encoding on the order sales text word vector sequence to obtain a second-scale feature vector; the first-scale feature vector and the second-scale feature vector are cascaded to obtain the order sales text semantic feature vector.

[0052] In the above-mentioned e-commerce platform order data intelligent analysis system 100, the user satisfaction analysis report generation module 130 is used to generate a user satisfaction analysis report based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector. It should be understood that the optimized order evaluation multimodal semantic feature vector contains a variety of information expressed by customers in the evaluation, such as text descriptions, pictures, and ratings. These vectors can more comprehensively capture customers' feelings and feedback on products or services by integrating text and visual information. Combined with the order sales text semantic feature vector, the customer's experience and expectations during the purchase process can be comprehensively analyzed, thereby revealing the key factors affecting customer satisfaction. The obtained user satisfaction analysis report can support marketing and customer relationship management decisions. Based on in-depth data analysis, companies can accurately identify the needs and preferences of different customer groups and formulate personalized marketing strategies and customer communication plans.

[0053] Figure 4 Schematic diagram of a block diagram of a user satisfaction analysis report generation module in an e-commerce platform order data intelligent analysis system according to an embodiment of the present application. Figure 4 As shown, in a specific embodiment of the present application, the user satisfaction analysis report generation module 130 includes: an order data feature fusion unit 131, used to fuse the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector to obtain an order satisfaction feature vector; a satisfaction analysis generation unit 132, used to pass the order satisfaction feature vector through a generator to generate a user satisfaction analysis report.

[0054] It should be understood that the feature vector that integrates order evaluation and sales data can also provide a cross-channel and full life cycle customer perspective. By comprehensively analyzing data at different stages, companies can identify changes and demand evolution of customers throughout the purchase process, and then optimize product design, service delivery and customer support strategies. This comprehensive perspective helps companies achieve comprehensive and personalized customer relationship management and improve customer satisfaction and loyalty.

[0055] Furthermore, the generator can transform the order satisfaction feature vector into report content with practical business significance. Through deep learning models or natural language processing technology, the generator can analyze various data features in the feature vector and generate structured report documents according to preset templates or rules. These reports not only include an overall assessment of customer satisfaction, but also in-depth discussions on the changing trends of various indicators, correlation analysis, and priority ranking of influencing factors, providing enterprises with clear data-driven decision support.

[0056] In summary, the embodiment of the present application first obtains the evaluation data and order sales data of the e-commerce platform orders, and then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally generates a user satisfaction analysis report through a generator, thereby improving customer experience and helping companies improve product and service quality.

[0057] As described above, the e-commerce platform order data intelligent analysis system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the e-commerce platform order data intelligent analysis system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the e-commerce platform order data intelligent analysis system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the e-commerce platform order data intelligent analysis system 100 can also be one of the many hardware modules of the terminal device.

[0058] Alternatively, in another example, the e-commerce platform order data intelligent analysis system 100 and the terminal device may also be separate devices, and the e-commerce platform order data intelligent analysis system 100 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0059] Figure 5 Flow chart of the method for intelligent analysis of e-commerce platform order data according to an embodiment of the present application. Figure 5 As shown, according to the e-commerce platform order data intelligent analysis method of the embodiment of the present application, it includes: S110, obtaining the evaluation data and order sales data of the e-commerce platform order; S120, extracting the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector from the evaluation data of the e-commerce platform order and the order sales data; S130, generating a user satisfaction analysis report based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector.

[0060] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned e-commerce platform order data intelligent analysis method have been referred to above. Figures 1 to 4 The invention has been introduced in detail in the description of the e-commerce platform order data intelligent analysis system, and therefore, its repeated description will be omitted.

[0061] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0062] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0064] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0065] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0066] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit of the technical solution of the present invention.

Claims

1. An e-commerce platform order data intelligent analysis system, characterized in that: include: The e-commerce platform order data acquisition module is used to obtain the evaluation data and order sales data of the e-commerce platform orders; An e-commerce platform order data extraction module, used to extract optimized order evaluation multimodal semantic feature vectors and order sales text semantic feature vectors from the evaluation data of the e-commerce platform orders and the order sales data; A user satisfaction analysis report generating module is used to generate a user satisfaction analysis report based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector.

2. The e-commerce platform order data intelligent analysis system according to claim 1 is characterized in that: The e-commerce platform order data extraction module includes: A customer service evaluation feature extraction unit, used to extract customer service evaluation features from the evaluation data of the e-commerce platform order to obtain a customer service evaluation semantic feature vector; A picture-sharing end evaluation feature extraction unit is used to extract a picture-sharing end evaluation feature from the evaluation data of the e-commerce platform order to obtain a picture-sharing end evaluation two-dimensional semantic feature vector; An order evaluation multimodal construction unit, used to construct the customer service end evaluation semantic feature vector and the image sharing end evaluation two-dimensional semantic feature vector to obtain the optimized order evaluation multimodal semantic feature vector; The order sales text feature extraction unit is used to extract features from the order sales data to obtain a semantic feature vector of the order sales text.

3. The e-commerce platform order data intelligent analysis system according to claim 2 is characterized in that: The client evaluation feature extraction unit includes: Extracting customer service evaluation data from the evaluation data of the e-commerce platform order; Segmenting the customer service evaluation data to obtain a customer service evaluation text sequence; The customer service evaluation text sequence is passed through a customer service evaluation bidirectional long short-term memory neural network model to obtain the customer service evaluation semantic feature vector.

4. The e-commerce platform order data intelligent analysis system according to claim 3 is characterized in that: The image printing end evaluation feature extraction unit comprises: A picture sharing end evaluation data extraction subunit, used to extract the picture sharing end evaluation data from the evaluation data of the e-commerce platform order; A picture-sharing end evaluation semantic coding subunit, used for semantically coding the picture-sharing end evaluation data to obtain a two-dimensional text feature matrix of the picture-sharing end evaluation; The picture-printing end evaluation two-dimensional convolutional encoding subunit is used to obtain the picture-printing end evaluation two-dimensional semantic feature vector by passing the picture-printing end evaluation two-dimensional text feature matrix through the picture-printing end evaluation two-dimensional text word granularity feature extractor.

5. The e-commerce platform order data intelligent analysis system according to claim 4 is characterized in that: The image-sharing end evaluation semantic coding subunit includes: The extracted picture-printing end evaluation data is passed through a picture-printing end evaluation text context encoder to obtain a plurality of picture-printing end evaluation feature vectors; The plurality of picture-printing-end evaluation feature vectors are arranged in two dimensions to obtain the picture-printing-end evaluation two-dimensional text feature matrix.

6. The e-commerce platform order data intelligent analysis system according to claim 5 is characterized in that: The order evaluation multimodal construction unit includes: Associating the customer service end evaluation semantic feature vector with the picture sharing end evaluation two-dimensional semantic feature vector to obtain an order evaluation multimodal semantic feature vector; Based on the order sales text semantic feature vector, the order evaluation multimodal semantic feature vector is optimized by a simulated transformer structure based on fine-grained distribution difference to obtain an optimized order evaluation multimodal semantic feature vector.

7. The e-commerce platform order data intelligent analysis system according to claim 6, characterized in that: Based on the order sales text semantic feature vector, the order evaluation multimodal semantic feature vector is optimized by a simulated transformer structure based on fine-grained distribution difference to obtain an optimized order evaluation multimodal semantic feature vector, including: Calculating the position difference between the multimodal semantic feature vector of the order evaluation and the semantic feature vector of the order sales text to obtain a position difference representation vector of the e-commerce platform order shift; After concatenating the order evaluation multimodal semantic feature vector and the order sales text semantic feature vector into an e-commerce platform order joint feature vector, the e-commerce platform order joint feature vector is input into a one-dimensional convolutional layer to obtain an e-commerce platform order shift convolutional coding representation vector; After performing e-commerce platform order linear embedding modulation encoding on the order evaluation multimodal semantic feature vector using the first e-commerce platform order linear embedding weight matrix to obtain an order evaluation multimodal semantic feature vector linear modulation vector, a position difference between the order evaluation multimodal semantic feature vector linear modulation vector and the order sales text semantic feature vector is calculated to obtain an e-commerce platform order shift modulation difference representation vector; Taking the shifted position differential representation vector of the e-commerce platform order as the value vector, the shifted convolutional coding representation vector of the e-commerce platform order and the shift modulated differential representation vector of the e-commerce platform order as the query vector and the key vector, the shifted position differential representation vector of the e-commerce platform order, the shifted convolutional coding representation vector of the e-commerce platform order and the shift modulated differential representation vector of the e-commerce platform order are input into a shift corrector based on an imitation transformer structure to obtain a multi-scale hierarchical fusion shift modulation weight vector of the e-commerce platform order; The optimized order evaluation multimodal semantic feature vector is obtained by calculating the point-by-point multiplication between the multi-scale hierarchical fusion shift modulation weight vector of the e-commerce platform order and the multimodal semantic feature vector of the order evaluation.

8. The e-commerce platform order data intelligent analysis system according to claim 7, characterized in that: The order sales text feature extraction unit comprises: The order sales data is passed through an order sales data word embedding encoder to obtain an order sales text word vector sequence; The order sales text word vector sequence is passed through an order sales convolutional neural network based on a multi-scale neighborhood feature extraction module to obtain the order sales text semantic feature vector.

9. The e-commerce platform order data intelligent analysis system according to claim 8, characterized in that: The user satisfaction analysis report generating module includes: An order data feature fusion unit, used to fuse the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector to obtain an order satisfaction feature vector; The satisfaction analysis generating unit is used to pass the order satisfaction feature vector through a generator to generate a user satisfaction analysis report.

10. An intelligent analysis method for order data of an e-commerce platform, characterized in that: include: Obtain evaluation data and order sales data of e-commerce platform orders; Extracting optimized order evaluation multimodal semantic feature vectors and order sales text semantic feature vectors from the evaluation data of the e-commerce platform orders and the order sales data; A user satisfaction analysis report is generated based on the optimized order evaluation multimodal semantic feature vector and the order sales text semantic feature vector.

Images