A digital commodity transaction intelligent management platform and method

Abstract

The present application provides a digital commodity transaction intelligent management platform and method, which relates to the field of intelligent transaction management. It performs semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain the digital commodity demand semantic features, and performs semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain the digital commodity supply semantic features, and then matches the digital commodity demand semantic features with the digital commodity supply semantic features in the same feature space, and finally classifies the matching features to determine whether to push the digital commodity supplier to be recommended to the digital commodity demander. In this way, the deep meaning and contextual relationship in the digital commodity information can be fully captured, so that the supply and demand sides can be accurately matched, thereby improving transaction efficiency and user experience.

Description

Technical Field

[0001] The present application relates to the field of intelligent transaction management, and more specifically, to an intelligent management platform and method for digital commodity transactions. Background Art

[0002] In today's era of rapid digital development, digital goods have become an indispensable part of people's lives. With the rapid advancement of information technology and the widespread popularity of the Internet, the exchange of digital goods in virtual environments has become increasingly frequent. Digital goods refer to physical products that are exchanged in a virtual environment. They exist in digital format and are transmitted through digital networks. These products cover a wide range, including but not limited to e-books, which provide readers with a convenient reading experience and allow them to enjoy a feast of knowledge anytime and anywhere; various accounts to meet the usage needs of different users on specific platforms; electronic vouchers to bring benefits and convenience to consumers; virtual currency or items in games to enrich players' gaming experience in the virtual world.

[0003] However, in the traditional digital commodity trading market, information matching between supply and demand often relies on simple keyword searches or manual screening. This method is not only inefficient, but also requires a lot of time and effort to search and compare one by one. It is also easy to miss potential matches, making it difficult for some high-quality digital commodities to be discovered by demanders in a timely manner. It may also make it difficult for suppliers to find suitable buyers for their commodities, greatly limiting the development vitality and resource allocation efficiency of the digital commodity trading market.

[0004] Therefore, an intelligent management solution for digital commodity transactions is needed. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a digital commodity transaction intelligent management platform and method, which performs semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain the digital commodity demand semantic features, and performs semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain the digital commodity supply semantic features, and then matches the digital commodity demand semantic features and the digital commodity supply semantic features in the same feature space, and finally classifies the matching features to determine whether to push the digital commodity supplier to be recommended to the digital commodity demander. In this way, the deep meaning and contextual relationship in the digital commodity information can be fully captured, so that the supply and demand sides can be accurately connected, thereby improving transaction efficiency and enhancing user experience.

[0006] According to one aspect of the present application, there is provided a digital commodity transaction intelligent management platform, which includes:

[0007] A digital commodity transaction-related information collection module, which is used to obtain the commodity information of the target digital commodity released by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended;

[0008] A digital commodity demand information semantic encoding module, used to perform semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain a digital commodity demand semantic feature vector;

[0009] A digital commodity supply information semantic encoding module, used for performing semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector;

[0010] A digital commodity demand-supply feature association module, configured to multiply the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix;

[0011] A digital commodity demand-supply feature enhancement module, used to perform feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector;

[0012] The recommendation result generating module is used to obtain a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander based on the digital commodity demand-supply association matching global semantic feature vector.

[0013] In the above-mentioned digital commodity transaction intelligent management platform, the digital commodity demand information semantic encoding module includes: a digital commodity demand information related word embedding unit, which is used to segment the commodity information of the target digital commodity released by the digital commodity demander and pass it through the word embedding layer to obtain a sequence of digital commodity demand information related word embedding vectors; a digital commodity demand semantic feature generation unit, which is used to arrange the sequence of digital commodity demand information related word embedding vectors in two dimensions into a two-dimensional input matrix of digital commodity demand related words and then pass it through a digital commodity demand semantic feature analyzer to obtain the digital commodity demand semantic feature vector.

[0014] In the above-mentioned digital commodity transaction intelligent management platform, the digital commodity supply information semantic encoding module includes: a digital commodity asset information related word embedding unit, which is used to segment the digital commodity asset information of the digital commodity supplier to be recommended and pass it through the word embedding layer to obtain a sequence of digital commodity asset information related word embedding vectors; a digital commodity supply semantic feature generation unit, which is used to arrange the sequence of digital commodity asset information related word embedding vectors in two dimensions into a two-dimensional input matrix of digital commodity asset related words and then pass it through a digital commodity supply semantic feature analyzer to obtain the digital commodity supply semantic feature vector.

[0015] In the above-mentioned digital commodity transaction intelligent management platform, the digital commodity demand-supply feature enhancement module includes: a digital commodity demand-supply association matching feature matrix segmentation unit, which is used to segment the digital commodity demand-supply association matching feature matrix to obtain multiple digital commodity demand-supply association matching local feature sub-matrices, wherein the sizes of the multiple digital commodity demand-supply association matching local feature sub-matrices are equal; a digital commodity demand-supply association matching feature sub-matrix expansion unit, which is used to expand the multiple digital commodity demand-supply association matching local feature sub-matrices to obtain multiple digital commodity demand-supply association matching local feature vectors; a digital commodity demand-supply association matching global semantic feature generation unit, which is used to pass the multiple digital commodity demand-supply association matching local feature vectors through a digital commodity association matching context semantic feature encoder to obtain the digital commodity demand-supply association matching global semantic feature vector.

[0016] In the above-mentioned digital commodity transaction intelligent management platform, the digital commodity demand semantic feature analyzer and the digital commodity supply semantic feature analyzer are convolutional neural network models using two-dimensional convolution kernels, and the digital commodity association matching context semantic feature encoder is a converter-based context semantic encoder.

[0017] In the above-mentioned digital commodity transaction intelligent management platform, the recommendation result generation module is used to: pass the optimized digital commodity demand-supply association matching global semantic feature vector through a classifier-based recommendation result generator to obtain a classification result, and the classification result is used to indicate whether to push the digital commodity supplier to be recommended to the digital commodity demander.

[0018] In the above-mentioned digital commodity transaction intelligent management platform, the recommendation result generation module includes: an association matching global semantic feature optimization unit, which is used to perform displacement compensation on the digital commodity demand-supply association matching global semantic feature vector based on feature local displacement to obtain an optimized digital commodity demand-supply association matching global semantic feature vector; an optimization association matching global semantic feature parsing unit, which is used to pass the optimized digital commodity demand-supply association matching global semantic feature vector through a classifier-based recommendation result generator to obtain a classification result, and the classification result is used to indicate whether the digital commodity supplier to be recommended is pushed to the digital commodity demander.

[0019] In the above-mentioned digital commodity transaction intelligent management platform, the association matching global semantic feature optimization unit is used to: determine the first classification weight matrix and the second classification weight matrix of the classifier before and after each iterative update; matrix multiply the first classification weight matrix and the second classification weight matrix with the digital commodity demand-supply association matching global semantic feature vector respectively to obtain the first digital commodity demand-supply association matching modulation feature vector and the second digital commodity demand-supply association matching modulation feature vector; calculate the position difference between the first digital commodity demand-supply association matching modulation feature vector and the second digital commodity demand-supply association matching modulation feature vector to obtain the digital commodity demand-supply association matching shift information representation vector; calculate the first digital commodity demand-supply association matching modulation feature vector The F norm of the positional mean vector between the quantity and the second digital commodity demand-supply association matching modulation feature vector is used as a shift compensation scaling factor; the digital commodity demand-supply association matching shift information representation vector is linearly scaled by the shift compensation scaling factor to obtain a scaled digital commodity demand-supply association matching shift information representation vector, and the scaled digital commodity demand-supply association matching shift information representation vector is input into a Sigmoid activation function to obtain a digital commodity demand-supply association matching back-propagation shift compensation representation vector; the positional point multiplication of the digital commodity demand-supply association matching back-propagation shift compensation representation vector and the digital commodity demand-supply association matching global semantic feature vector is calculated to obtain the optimized digital commodity demand-supply association matching global semantic feature vector.

[0020] According to another aspect of the present application, a method for intelligent management of digital commodity transactions is provided, which includes:

[0021] Obtaining product information of target digital products published by digital product demanders and digital product asset information of digital product suppliers to be recommended;

[0022] Performing semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain a digital commodity demand semantic feature vector;

[0023] Performing semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector;

[0024] Multiplying the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix;

[0025] Performing feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector;

[0026] Based on the digital commodity demand-supply association matching global semantic feature vector, a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander is obtained.

[0027] In the above-mentioned intelligent management method for digital commodity transactions, feature enhancement is performed on the digital commodity demand-supply association matching feature matrix to obtain a global semantic feature vector of digital commodity demand-supply association matching, including: feature matrix segmentation of the digital commodity demand-supply association matching feature matrix to obtain multiple digital commodity demand-supply association matching local feature sub-matrices, wherein the sizes of the multiple digital commodity demand-supply association matching local feature sub-matrices are equal; the multiple digital commodity demand-supply association matching local feature sub-matrices are expanded to obtain multiple digital commodity demand-supply association matching local feature vectors; the multiple digital commodity demand-supply association matching local feature vectors are passed through a digital commodity association matching context semantic feature encoder to obtain the digital commodity demand-supply association matching global semantic feature vector.

[0028] In the above-mentioned intelligent management method for digital commodity transactions, based on the global semantic feature vector of the digital commodity demand-supply association matching, a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander is obtained, including: performing displacement compensation on the global semantic feature vector of the digital commodity demand-supply association matching based on feature local displacement to obtain an optimized global semantic feature vector of the digital commodity demand-supply association matching; passing the optimized global semantic feature vector of the digital commodity demand-supply association matching through a recommendation result generator based on a classifier to obtain a classification result, and the classification result is used to indicate whether to push the digital commodity supplier to be recommended to the digital commodity demander.

[0029] Compared with the prior art, the digital commodity transaction intelligent management platform and method provided by the present application obtains digital commodity demand semantic features by performing semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander, and obtains digital commodity supply semantic features by performing semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended, and then matches the digital commodity demand semantic features with the digital commodity supply semantic features in the same feature space, and finally classifies the matching features to determine whether to push the digital commodity supplier to be recommended to the digital commodity demander. In this way, the deep meaning and contextual relationship in the digital commodity information can be fully captured, so that the supply and demand sides can be accurately matched, thereby improving transaction efficiency and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] 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.

[0031] Figure 1 4 is a block diagram of a digital commodity transaction intelligent management platform according to an embodiment of the present application.

[0032] Figure 2 4 is a block diagram of a digital commodity demand information semantic encoding module in a digital commodity transaction intelligent management platform according to an embodiment of the present application.

[0033] Figure 3 4 is a block diagram of a digital commodity supply information semantic encoding module in a digital commodity transaction intelligent management platform according to an embodiment of the present application.

[0034] Figure 4 4 is a block diagram of a digital commodity demand-supply feature enhancement module in a digital commodity transaction intelligent management platform according to an embodiment of the present application.

[0035] Figure 5 This is a block diagram of a recommendation result generation module in a digital commodity transaction intelligent management platform according to an embodiment of the present application.

[0036] Figure 6 Flow chart of a digital commodity transaction intelligent management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0037] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0038] As mentioned in the background technology above, digital goods refer to physical products that are exchanged in a virtual environment. They exist in digital format and are transmitted through digital networks. These products include but are not limited to e-books, accounts, electronic vouchers, virtual currency or items in games, etc. In the traditional digital commodity trading market, information matching between supply and demand often relies on simple keyword searches or manual screening, which is inefficient and prone to missing potential matches. Therefore, an intelligent management solution for digital commodity transactions is expected.

[0039] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for the intelligent management of digital commodity transactions.

[0040] Figure 1 FIG. 1 is a block diagram of a digital commodity transaction intelligent management platform according to an embodiment of the present application. Figure 1 As shown, in the digital commodity transaction intelligent management platform 100, it includes: a digital commodity transaction related information collection module 110, which is used to obtain the commodity information of the target digital commodity released by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended; a digital commodity demand information semantic encoding module 120, which is used to perform semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain a digital commodity demand semantic feature vector; a digital commodity supply information semantic encoding module 130, which is used to perform semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector; a digital commodity demand information semantic encoding module 130, which is used to perform semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector; The demand-supply feature association module 140 is used to multiply the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix; the digital commodity demand-supply feature enhancement module 150 is used to perform feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector; the recommendation result generation module 160 is used to obtain a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander based on the digital commodity demand-supply association matching global semantic feature vector.

[0041] Specifically, in the technical solution of the present application, firstly, the commodity information of the target digital commodity published by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended are obtained. It should be understood that the commodity information of the target digital commodity published by the digital commodity demander specifically includes the name of the digital commodity, the digital commodity description including the commodity function, purpose and other information, the commodity attributes, and the requirements including the acceptable price range, quantity, and delivery method of the digital commodity; the digital commodity asset information of the digital commodity supplier to be recommended includes the list information of the tradable digital commodities owned, including the basic information such as the commodity name and introduction, in addition to which, detailed commodity information such as detailed description, attributes, price, inventory and related service commitment information are also included. In general, by performing semantic analysis on the commodity information of the target digital commodity published by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended, the respective semantic features are extracted, and then matching is performed in the same feature space. This matching process directly determines whether the commodity of the supplier is consistent with the demand of the demander. In particular, in a specific implementable embodiment of the present application, the commodity information of the target digital commodity released by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended can be obtained from the database of the digital commodity transaction intelligent management platform.

[0042] Figure 2 FIG. 1 is a block diagram of a digital commodity demand information semantic coding module in a digital commodity transaction intelligent management platform according to an embodiment of the present application. Figure 2 As shown, the digital commodity demand information semantic encoding module 120 includes: a digital commodity demand information related word embedding unit 121, which is used to segment the commodity information of the target digital commodity released by the digital commodity demander and pass it through a word embedding layer to obtain a sequence of digital commodity demand information related word embedding vectors; a digital commodity demand semantic feature generating unit 122, which is used to arrange the sequence of digital commodity demand information related word embedding vectors in two dimensions into a two-dimensional input matrix of digital commodity demand related words and then pass it through a digital commodity demand semantic feature analyzer to obtain the digital commodity demand semantic feature vector.

[0043] Considering that directly using the commodity information of the target digital commodity as a processing unit will greatly increase the computational complexity and storage space requirements, in order to reduce the amount of data during processing and improve processing efficiency, the commodity information of the target digital commodity published by the digital commodity demander can be segmented. The word segmentation process can divide continuous text characters into independent vocabulary units, so that the computer can better understand the commodity information of the digital commodity when processing the text, which helps to distinguish the different attributes and characteristics of the commodity and provide a basis for subsequent semantic analysis. Then the commodity information of the target digital commodity after word segmentation is passed through the word embedding layer to obtain a sequence of word embedding vectors related to the digital commodity demand information. Through the word embedding layer, each word in the commodity information after word segmentation is converted into a vector containing rich semantic information. It should be understood that if the word after word segmentation is directly used as a processing unit, it will face problems such as high-dimensional sparsity and high computational complexity. The word embedding vector is low-dimensional and dense, which can be easy to calculate while retaining sufficient semantic information.

[0044] In order to preserve the spatial relationship between word embedding vectors, thereby reflecting the semantic connection between words to a certain extent, it is necessary to arrange the sequence of word embedding vectors related to the digital commodity demand information in two dimensions into a two-dimensional input matrix of digital commodity demand related words. Considering that in order to capture the local features in the two-dimensional input matrix of digital commodity demand related words, and then extract deeper semantic information, so as to truly understand the content of digital commodity demand information, it is necessary to pass the two-dimensional input matrix of digital commodity demand related words through a digital commodity demand semantic feature analyzer to obtain a digital commodity demand semantic feature vector. In this application, the digital commodity demand semantic feature analyzer is a convolutional neural network model using a two-dimensional convolution kernel. The two-dimensional convolution kernel can slide on the input matrix, and capture the local semantic patterns between words through convolution operations. These patterns represent specific semantic units or concepts, and as the number of convolution layers increases, the convolutional neural network model can extract more and more advanced and abstract semantic features. These features not only contain information at the vocabulary level, but also contain information at the sentence, paragraph, and even the entire text level, which helps to more comprehensively understand the content of digital commodity demand information.

[0045] Figure 3 FIG. 1 is a block diagram of a digital commodity supply information semantic coding module in a digital commodity transaction intelligent management platform according to an embodiment of the present application. Figure 3As shown, the digital commodity supply information semantic encoding module 130 includes: a digital commodity asset information related word embedding unit 131, which is used to segment the digital commodity asset information of the digital commodity supplier to be recommended and pass it through the word embedding layer to obtain a sequence of digital commodity asset information related word embedding vectors; a digital commodity supply semantic feature generation unit 132, which is used to arrange the sequence of digital commodity asset information related word embedding vectors in two dimensions into a digital commodity asset related word two-dimensional input matrix and then pass it through a digital commodity supply semantic feature analyzer to obtain the digital commodity supply semantic feature vector.

[0046] In order to help distinguish different digital commodity attributes, specifications, descriptions, etc. in the digital commodity asset information, it is necessary to perform word segmentation on the digital commodity asset information of the supplier of the digital commodity to be recommended. The digital commodity asset information after word segmentation is more convenient for semantic analysis. In order to more accurately identify key information when understanding the digital commodity asset information, the digital commodity asset information after word segmentation can be passed through the word embedding layer to obtain a sequence of word embedding vectors related to the digital commodity asset information. The word embedding vector can capture the similarity and correlation between words, help to deeply understand the connotation of digital commodity asset information, improve the understanding and processing capabilities of digital commodity asset information, and thus provide strong support for subsequent accurate recommendation services.

[0047] Next, the sequence of the digital commodity asset information related word embedding vectors is arranged in two dimensions into a two-dimensional input matrix of digital commodity asset related words, and then the digital commodity supply semantic feature analyzer is passed to obtain the digital commodity supply semantic feature vector. In the present application, the digital commodity supply semantic feature analyzer is a convolutional neural network model using a two-dimensional convolution kernel. It should be understood that there are contextual relationships and semantic connections between the words in the digital commodity asset information. In order to retain these contextual semantic relationships, the sequence of digital commodity asset information related word embedding vectors can be arranged in two dimensions into a two-dimensional input matrix of digital commodity asset related words. The format of the two-dimensional input matrix also meets the input requirements of the convolutional neural network model. The convolution layer in the convolutional neural network model can extract robust features by sharing weights and local connections. These features not only contain information at the vocabulary level of digital commodity asset information related words, but also contain contextual relationships and semantic connections between words, which is conducive to improving the accuracy of digital commodity asset feature extraction.

[0048] Figure 4 FIG. 1 is a block diagram of a digital commodity demand-supply feature enhancement module in a digital commodity transaction intelligent management platform according to an embodiment of the present application. Figure 4As shown, the digital commodity demand-supply feature enhancement module 150 includes: a digital commodity demand-supply association matching feature matrix segmentation unit 151, which is used to segment the digital commodity demand-supply association matching feature matrix to obtain multiple digital commodity demand-supply association matching local feature sub-matrices, wherein the sizes of the multiple digital commodity demand-supply association matching local feature sub-matrices are equal; a digital commodity demand-supply association matching feature sub-matrix expansion unit 152, which is used to expand the multiple digital commodity demand-supply association matching local feature sub-matrices to obtain multiple digital commodity demand-supply association matching local feature vectors; a digital commodity demand-supply association matching global semantic feature generation unit 153, which is used to pass the multiple digital commodity demand-supply association matching local feature vectors through a digital commodity association matching context semantic feature encoder to obtain the digital commodity demand-supply association matching global semantic feature vector.

[0049] In order to transform the digital commodity supply semantic feature vector and the digital commodity demand semantic feature vector from the original space to a new common space, it is necessary to multiply the transposed vector of the digital commodity demand semantic feature vector and the digital commodity supply semantic feature vector to obtain the digital commodity demand-supply association matching feature matrix. The matching feature matrix obtained by multiplication can quantitatively represent the matching degree between digital commodity demand and supply, and then decide whether to recommend the digital commodity supplier to be recommended to the digital commodity demander.

[0050] In an embodiment of the present application, a possible implementation method of multiplying the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix may be: associating the digital commodity demand semantic feature vector and the digital commodity supply semantic feature vector with the following association coding formula to obtain a digital commodity demand-supply association matching feature matrix; wherein the association coding formula is:

[0051]

[0052] Where V a represents the digital commodity supply semantic feature vector, V T a represents the transposed vector of the digital commodity supply semantic feature vector, V b represents the digital commodity demand semantic feature vector, M represents the digital commodity demand-supply association matching feature matrix, Represents vector multiplication.

[0053] The information content size of different digital commodity suppliers and digital commodity demanders is different. When the information content of the processed digital commodity suppliers and digital commodity demanders is relatively rich, the scale of the demand-supply association matching feature matrix will be relatively large. Directly processing the entire large matrix will result in excessive computational complexity and affect the response speed of the system. Therefore, in the technical solution of the present application, the digital commodity demand-supply association matching feature matrix is ​​divided into feature matrices to obtain multiple digital commodity demand-supply association matching local feature sub-matrices, wherein the sizes of the digital commodity demand-supply association matching local feature sub-matrices in the multiple digital commodity demand-supply association matching local feature sub-matrices are equal. By dividing into multiple small matrices, the complexity of a single calculation can be reduced and the processing efficiency can be improved. At the same time, different local feature sub-matrices may represent different demand-supply matching characteristics. By dividing, each sub-matrix can be analyzed and processed more refinedly, so as to more accurately capture the subtle differences and associations between demand and supply.

[0054] Next, the multiple digital commodity demand-supply association matching local feature submatrices are expanded to obtain multiple digital commodity demand-supply association matching local feature vectors. It should be understood that although the digital commodity demand-supply association matching local feature submatrix reduces the complexity of the overall matrix to a certain extent, in some processing flows, data in vector form is more intuitive and easier to process. By expanding the submatrix into a vector, the subsequent data processing steps can be simplified and the execution efficiency of the algorithm can be improved.

[0055] In order to capture the complex dependencies between multiple local feature vectors of digital commodity demand-supply association matching, so as to help deeply understand the intrinsic connection between demand and supply, and then more accurately judge the matching degree between the two, it is necessary to pass the multiple local feature vectors of digital commodity demand-supply association matching through the digital commodity association matching context semantic feature encoder to obtain the global semantic feature vector of digital commodity demand-supply association matching. In this application, the digital commodity association matching context semantic feature encoder is a converter-based context semantic encoder. The converter-based context semantic encoder can capture the dependencies between different association matching local feature vectors through the self-attention mechanism and generate a representation rich in context information. The global semantic feature vector containing rich context information can more accurately reflect the matching relationship between digital commodity demand and supply, thereby improving the accuracy of matching. When judging whether to push the supplier of the digital commodity to be recommended to the demander, a more accurate matching result means a lower mispush rate and higher user satisfaction.

[0056] Figure 5 FIG. 1 is a block diagram of a recommendation result generation module in a digital commodity transaction intelligent management platform according to an embodiment of the present application. Figure 5As shown, the recommendation result generation module 160 includes: an association matching global semantic feature optimization unit 161, which is used to perform displacement compensation on the digital commodity demand-supply association matching global semantic feature vector based on feature local displacement to obtain an optimized digital commodity demand-supply association matching global semantic feature vector; an optimization association matching global semantic feature parsing unit 162, which is used to pass the optimized digital commodity demand-supply association matching global semantic feature vector through a classifier-based recommendation result generator to obtain a classification result, and the classification result is used to indicate whether to push the digital commodity supplier to be recommended to the digital commodity demander.

[0057] In particular, considering that the commodity information and digital commodity asset information of the target digital commodity obtained in actual applications often contain noise, such as synonyms and different expressions, the presence of these noises will lead to inaccurate extraction of local features of the extracted digital commodity demand semantic feature vector and digital commodity supply semantic feature vector, thereby affecting the subsequent construction of the global semantic feature vector of digital commodity demand-supply association matching, that is, the disturbance of local features in the semantic feature vector will affect the performance of the global semantic feature. When the classifier is not robust enough to the disturbance of local feature input, it may lead to erroneous classification results, thereby affecting the accuracy of the recommendation, especially in the dynamic and changing digital commodity trading market environment, user needs and commodity information will continue to change. In order to better adapt to these changes and reduce the impact of local feature input disturbance on the recommendation results, therefore, in the technical solution of the present application, it is necessary to perform displacement compensation based on feature local displacement on the global semantic feature vector of the digital commodity demand-supply association matching to obtain an optimized global semantic feature vector of digital commodity demand-supply association matching.

[0058] Specifically, in an embodiment of the present application, the global semantic feature vector of the digital commodity demand-supply association matching is subjected to displacement compensation based on feature local displacement to obtain an optimized global semantic feature vector of the digital commodity demand-supply association matching, including: determining a first classification weight matrix and a second classification weight matrix of the classifier before and after each iterative update; performing matrix multiplication of the first classification weight matrix and the second classification weight matrix with the global semantic feature vector of the digital commodity demand-supply association matching respectively to obtain a first digital commodity demand-supply association matching modulation feature vector and a second digital commodity demand-supply association matching modulation feature vector; calculating the positional difference between the first digital commodity demand-supply association matching modulation feature vector and the second digital commodity demand-supply association matching modulation feature vector to obtain a digital commodity demand-supply association matching displacement information representation vector; calculating Calculate the F norm of the position mean vector between the first digital commodity demand-supply association matching modulation feature vector and the second digital commodity demand-supply association matching modulation feature vector as a shift compensation scaling factor; linearly scale the digital commodity demand-supply association matching shift information representation vector by the shift compensation scaling factor to obtain a scaled digital commodity demand-supply association matching shift information representation vector, and input the scaled digital commodity demand-supply association matching shift information representation vector into a Sigmoid activation function to obtain a digital commodity demand-supply association matching back-propagation shift compensation representation vector; calculate the position point multiplication of the digital commodity demand-supply association matching back-propagation shift compensation representation vector and the digital commodity demand-supply association matching global semantic feature vector to obtain the optimized digital commodity demand-supply association matching global semantic feature vector.

[0059] In the implementation of the present application, another expression method of performing displacement compensation based on feature local displacement on the global semantic feature vector of the digital commodity demand-supply association matching to obtain an optimized global semantic feature vector of the digital commodity demand-supply association matching can be: processing the global semantic feature vector of the digital commodity demand-supply association matching with the following formula to obtain an optimized global semantic feature vector of the digital commodity demand-supply association matching; wherein the formula is:

[0060]

[0061] Wherein, M1 is the first classification weight matrix, M2 is the second classification weight matrix, V is the global semantic feature vector of the digital commodity demand-supply association matching, represents matrix multiplication, Indicates adding by position point, ‖·‖ F represents the F norm of the feature vector, S represents the shift compensation scaling factor, ⊙ represents the point multiplication by position, represents point-by-point subtraction, sigmoid represents an activation function, and V' represents the optimized global semantic feature vector of the digital commodity demand-supply association matching.

[0062] In order to enhance the robustness of the classifier to local disturbances in the input feature vector, the global semantic feature vector of the digital commodity demand-supply association matching is compensated for the displacement based on the back propagation representation. The key to the displacement compensation based on the back propagation representation of the global semantic feature vector of the digital commodity demand-supply association matching is to capture the information of the classification weight matrix changing with the training process, and to use this information to generate the digital commodity demand-supply association matching displacement information representation vector, and then generate the digital commodity demand-supply association matching back propagation displacement compensation representation vector through scaling and nonlinear transformation (Sigmoid function). In this way, the local displacement of the feature of the global semantic feature vector of the digital commodity demand-supply association matching is described by supporting the distribution difference of different classification scenarios of the classifier during the training process, so that not only the displacement of the feature vector caused by the change of the classification weight matrix can be quantified, but also a compensation vector can be generated according to the displacement information. The role of the compensation vector is to weight and adjust each element in the feature vector according to the degree of influence of the disturbance, thereby reducing the influence of the disturbance on the classification result. In this way, the robustness of the classifier to local disturbances in the input data can be significantly improved, the stability and generalization ability of the classifier can be enhanced, and by reducing the impact of local disturbances on the classification results, the classifier can maintain a high classification accuracy when facing noisy or variant data.

[0063] Finally, the optimized digital commodity demand-supply association matching global semantic feature vector is passed through the classifier-based recommendation result generator to obtain a classification result, and the classification result is used to indicate whether to push the digital commodity supplier to be recommended to the digital commodity demander. It should be understood that although the optimized digital commodity demand-supply association matching global semantic feature vector contains rich context information and matching relationships, it does not directly give a specific conclusion on whether to recommend. Through the classifier-based recommendation result generator, these complex feature information can be converted into clear classification results (such as "recommended" or "not recommended"), so as to clarify the decision boundary and facilitate the system to make the final recommendation decision. The classification result directly determines whether to push the digital commodity supplier to be recommended to the digital commodity demander. If the classification result is "recommended", the system will perform the push operation; if the classification result is "not recommended", the system will not push. This clear decision guidance helps to ensure the accuracy and effectiveness of the recommendation. Accurate classification results can ensure that the digital commodities that best meet the user's needs are recommended to the user, thereby improving the user's shopping experience and satisfaction.

[0064] In summary, the digital commodity transaction intelligent management platform 100 based on the embodiment of the present application is explained, which performs semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain the digital commodity demand semantic features, and performs semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain the digital commodity supply semantic features, and then matches the digital commodity demand semantic features and the digital commodity supply semantic features in the same feature space, and finally classifies the matching features to determine whether to push the digital commodity supplier to be recommended to the digital commodity demander. In this way, the deep meaning and contextual relationship in the digital commodity information can be fully captured, so that the supply and demand sides can be accurately connected, thereby improving transaction efficiency and enhancing user experience.

[0065] As described above, the digital commodity transaction intelligent management platform 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for intelligent management of digital commodity transactions. In one example, the digital commodity transaction intelligent management platform 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the digital commodity transaction intelligent management platform 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 digital commodity transaction intelligent management platform 100 can also be one of the many hardware modules of the terminal device.

[0066] Alternatively, in another example, the digital commodity transaction intelligent management platform 100 and the terminal device may also be separate devices, and the digital commodity transaction intelligent management platform 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.

[0067] Figure 6 Flowchart of the digital commodity transaction intelligent management method according to the embodiment of the present application. Figure 6As shown, in the digital commodity transaction intelligent management method, it includes: S110, obtaining the commodity information of the target digital commodity released by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended; S120, performing semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain a digital commodity demand semantic feature vector; S130, performing semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector; S140, multiplying the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix; S150, performing feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector; S160, based on the digital commodity demand-supply association matching global semantic feature vector, obtaining a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander.

[0068] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned digital commodity transaction intelligent management method have been referred to above. Figures 1 to 5 The digital commodity transaction intelligent management platform has been introduced in detail, and therefore, its repeated description will be omitted.

[0069] In summary, the digital commodity transaction intelligent management method based on the embodiment of the present application is explained, which performs semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain the digital commodity demand semantic features, and performs semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain the digital commodity supply semantic features, and then matches the digital commodity demand semantic features and the digital commodity supply semantic features in the same feature space, and finally classifies the matching features to determine whether to push the digital commodity supplier to be recommended to the digital commodity demander. In this way, the deep meaning and contextual relationship in the digital commodity information can be fully captured, so that the supply and demand sides can be accurately matched, thereby improving transaction efficiency and user experience.

Claims

1. A digital commodity transaction intelligent management platform, characterized in that: include: A digital commodity transaction-related information collection module, which is used to obtain the commodity information of the target digital commodity released by the digital commodity demander and the digital commodity asset information of the digital commodity supplier to be recommended; A digital commodity demand information semantic encoding module, used for performing semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain a digital commodity demand semantic feature vector; A digital commodity supply information semantic encoding module, used for performing semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector; A digital commodity demand-supply feature association module, configured to multiply the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix; A digital commodity demand-supply feature enhancement module, used to perform feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector; A recommendation result generating module, used for obtaining a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander based on the digital commodity demand-supply association matching global semantic feature vector; Wherein, the recommendation result generation module includes: An association matching global semantic feature optimization unit, used for performing displacement compensation on the digital commodity demand-supply association matching global semantic feature vector based on feature local displacement to obtain an optimized digital commodity demand-supply association matching global semantic feature vector; an optimized association matching global semantic feature parsing unit, used for passing the optimized digital commodity demand-supply association matching global semantic feature vector through a classifier-based recommendation result generator to obtain a classification result, wherein the classification result is used to indicate whether to push the digital commodity supplier to be recommended to the digital commodity demander; Wherein, the association matching global semantic feature optimization unit is used to: Determine a first classification weight matrix and a second classification weight matrix of the classifier before and after each iterative update; Performing matrix multiplication of the first classification weight matrix and the second classification weight matrix with the digital commodity demand-supply association matching global semantic feature vector respectively to obtain a first digital commodity demand-supply association matching modulation feature vector and a second digital commodity demand-supply association matching modulation feature vector; Calculating the position difference between the first digital commodity demand-supply association matching modulation feature vector and the second digital commodity demand-supply association matching modulation feature vector to obtain a digital commodity demand-supply association matching shift information representation vector; Calculating the F-norm of the position mean vector between the first digital commodity demand-supply association matching modulation feature vector and the second digital commodity demand-supply association matching modulation feature vector as a shift compensation scaling factor; Linearly scaling the digital commodity demand-supply association matching shift information representation vector by the shift compensation scaling factor to obtain a scaled digital commodity demand-supply association matching shift information representation vector, and inputting the scaled digital commodity demand-supply association matching shift information representation vector into a Sigmoid activation function to obtain a digital commodity demand-supply association matching back-propagation shift compensation representation vector; The optimized digital commodity demand-supply association matching global semantic feature vector is obtained by calculating the position point multiplication of the digital commodity demand-supply association matching back-propagation shift compensation representation vector and the digital commodity demand-supply association matching global semantic feature vector.

2. The digital commodity transaction intelligent management platform according to claim 1, characterized in that: The digital commodity demand information semantic encoding module includes: A digital commodity demand information related word embedding unit, used to segment the commodity information of the target digital commodity released by the digital commodity demander and pass it through a word embedding layer to obtain a sequence of digital commodity demand information related word embedding vectors; The digital commodity demand semantic feature generating unit is used to arrange the sequence of digital commodity demand information related word embedding vectors in two dimensions into a two-dimensional input matrix of digital commodity demand related words and then pass it through a digital commodity demand semantic feature analyzer to obtain the digital commodity demand semantic feature vector.

3. The digital commodity transaction intelligent management platform according to claim 2, characterized in that: The digital commodity supply information semantic encoding module includes: A digital commodity asset information related word embedding unit, used to segment the digital commodity asset information of the digital commodity supplier to be recommended and pass it through the word embedding layer to obtain a sequence of digital commodity asset information related word embedding vectors; The digital commodity supply semantic feature generation unit is used to arrange the sequence of digital commodity asset information related word embedding vectors in two dimensions into a digital commodity asset related word two-dimensional input matrix and then pass it through the digital commodity supply semantic feature analyzer to obtain the digital commodity supply semantic feature vector.

4. The digital commodity transaction intelligent management platform according to claim 3, characterized in that: The digital commodity demand-supply feature enhancement module includes: a digital commodity demand-supply association matching feature matrix segmentation unit, used for segmenting the digital commodity demand-supply association matching feature matrix to obtain a plurality of digital commodity demand-supply association matching local feature sub-matrices, wherein the sizes of the respective digital commodity demand-supply association matching local feature sub-matrices in the plurality of digital commodity demand-supply association matching local feature sub-matrices are equal; A digital commodity demand-supply association matching feature submatrix expansion unit, used to expand the plurality of digital commodity demand-supply association matching local feature submatrices to obtain a plurality of digital commodity demand-supply association matching local feature vectors; The digital commodity demand-supply association matching global semantic feature generating unit is used to obtain the digital commodity demand-supply association matching global semantic feature vector by passing the multiple digital commodity demand-supply association matching local feature vectors through the digital commodity association matching context semantic feature encoder.

5. The digital commodity transaction intelligent management platform according to claim 4, characterized in that: The digital commodity demand semantic feature analyzer and the digital commodity supply semantic feature analyzer are convolutional neural network models using two-dimensional convolution kernels, and the digital commodity association matching context semantic feature encoder is a converter-based context semantic encoder.

6. A digital commodity transaction intelligent management method, using the digital commodity transaction intelligent management platform according to claim 1, characterized in that: include: Obtain product information of target digital products published by digital product demanders and digital product asset information of digital product suppliers to be recommended; Performing semantic analysis on the commodity information of the target digital commodity released by the digital commodity demander to obtain a digital commodity demand semantic feature vector; Performing semantic analysis on the digital commodity asset information of the digital commodity supplier to be recommended to obtain a digital commodity supply semantic feature vector; Multiplying the digital commodity demand semantic feature vector and the transposed vector of the digital commodity supply semantic feature vector to obtain a digital commodity demand-supply association matching feature matrix; Performing feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector; Based on the digital commodity demand-supply association matching global semantic feature vector, a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander is obtained.

7. The digital commodity transaction intelligent management method according to claim 6, characterized in that: Performing feature enhancement on the digital commodity demand-supply association matching feature matrix to obtain a digital commodity demand-supply association matching global semantic feature vector includes: Performing feature matrix segmentation on the digital commodity demand-supply association matching feature matrix to obtain a plurality of digital commodity demand-supply association matching local feature sub-matrices, wherein the sizes of the respective digital commodity demand-supply association matching local feature sub-matrices in the plurality of digital commodity demand-supply association matching local feature sub-matrices are equal; Expanding the plurality of digital commodity demand-supply association matching local feature sub-matrices to obtain a plurality of digital commodity demand-supply association matching local feature vectors; The multiple digital commodity demand-supply association matching local feature vectors are passed through a digital commodity association matching context semantic feature encoder to obtain the digital commodity demand-supply association matching global semantic feature vector.

8. The digital commodity transaction intelligent management method according to claim 7, characterized in that: Based on the digital commodity demand-supply association matching global semantic feature vector, obtaining a recommendation result of whether to push the digital commodity supplier to be recommended to the digital commodity demander, including: Performing displacement compensation based on feature local displacement on the digital commodity demand-supply association matching global semantic feature vector to obtain an optimized digital commodity demand-supply association matching global semantic feature vector; The optimized digital commodity demand-supply association matching global semantic feature vector is passed through a classifier-based recommendation result generator to obtain a classification result, and the classification result is used to indicate whether to push the digital commodity supplier to be recommended to the digital commodity demander.

Images