Cross-class commodity recommendation method, device and equipment and readable storage medium

Abstract

The invention provides a cross-category commodity recommendation method, device and equipment and a readable storage medium, and the method comprises the steps: carrying out the processing of a training commodity through a first classification model and a second classification model, and obtaining a first training category and a second training category; determining a category association table according to the first training categories and the second training categories of the plurality of training commodities; querying a category association table according to the first category of the initial commodity to obtain a second category; and determining recommended commodities corresponding to the initial commodities from the commodities included in the second category. According to the method, the category association table is obtained by utilizing the category classification results of the commodities in different category systems, so that cross-category commodity recommendation can be realized, and the cross-category commodity recommendation effect is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a cross-category goods recommendation method, device, equipment and readable storage medium. BACKGROUND

[0002] The category system generally refers to the collection of interrelated and interdependent goods varieties formed after the work of classifying goods, assigning goods codes and compiling goods catalogs according to a specific classification purpose. Different manufacturers will often establish different category systems based on their own business, and the same goods will often be classified into different named categories under different category systems. For example, outdoor pants will usually be classified into the clothing category under the category system established by outdoor sports manufacturers, and will usually be classified into the pants category under the category system established by clothing manufacturers.

[0003] However, due to the inconsistency of the names of categories with the same meaning under different category systems, it is not possible to accurately match the categories under other category systems for goods recommendation according to the text or semantics when recommending goods across category systems. SUMMARY

[0004] Therefore, it is necessary to provide a cross-category goods recommendation method, device, equipment and readable storage medium to improve the accuracy of cross-category goods recommendation in view of the above technical problems.

[0005] In a first aspect, the present application provides a cross-category goods recommendation method, comprising:

[0006] processing the training goods by the first classification model and the second classification model to obtain the first training category of the training goods in the first category system and the second training category of the training goods in the second category system;

[0007] determining the category association table between the first category system and the second category system according to the first training category and the second training category of the plurality of training goods;

[0008] querying the category association table according to the first category of the initial goods in the first category system to obtain the second category associated with the first category in the second category system;

[0009] determining the recommended goods corresponding to the initial goods from the goods included in the second category.

[0010] As a feasible embodiment of the present application, the determination of the category association table between the first category system and the second category system according to the first training category and the second training category of the plurality of training goods comprises:

[0011] constructing an initial category association table including initial association degrees between categories in a first category system and categories in a second category system;

[0012] determining target training commodities from the plurality of training commodities for a first target category in the first category system and a second target category in the second category system, wherein the first training category of the target training commodities in the first category system is the first target category, and the second training category of the target training commodities in the second category system is the second target category;

[0013] updating the initial association degrees in the initial category association table according to a proportion of the target training commodities in the training commodities, to obtain the category association table.

[0014] As a feasible embodiment of the present application, the updating of the initial association degrees in the initial category association table according to the proportion of the target training commodities in the training commodities, to obtain the category association table, comprises:

[0015] updating the initial association degrees in the initial category association table according to a comparison result of the proportion of the target training commodities in the training commodities and a preset proportion threshold, to obtain the category association table.

[0016] As a feasible embodiment of the present application, the updating of the initial association degrees in the initial category association table according to the comparison result of the proportion of the target training commodities in the training commodities and the preset proportion threshold, to obtain the category association table, comprises:

[0017] in a case where the proportion of the target training commodities in the training commodities is greater than the preset proportion threshold, updating the initial association degrees in the initial category association table according to the proportion, to obtain the category association table;

[0018] in a case where the proportion of the target training commodities in the training commodities is less than or equal to the preset proportion threshold, updating the initial association degrees in the initial category association table to preset association degrees.

[0019] As a feasible embodiment of the present application, the constructing of the initial category association table comprises:

[0020] processing the training categories in the first category system and the second category system according to a semantic recognition model, to obtain category representation feature information corresponding to each of the training categories;

[0021] generating an initial association table between the first category system and the second category system according to similarities between the category representation feature information corresponding to each of the training categories.

[0022] As an embodiment of the present application, the initial association degree in the initial category association table is updated according to the proportion of the number of target training goods in the training goods, to obtain the category association table, which comprises:

[0023] Obtaining external knowledge information of the first category system and the second category system, the external knowledge information comprising at least one of category goods proportion information and category brand proportion information in the first category system and the second category system;

[0024] Updating the initial association degree in the initial category association table according to the external knowledge information and the proportion of the number of target training goods in the training goods, to obtain the category association table.

[0025] As an embodiment of the present application, before the first category in the first category system is used to query the category association table to obtain the second category associated with the first category in the second category system, the method further comprises:

[0026] Using the first category in the first category system to query a plurality of category association tables to obtain a reference category with the highest association degree with the first category in each candidate category system, respectively;

[0027] Determining a target category system from the candidate category systems according to the association degree between the first category system and each candidate category system and the association degree between the reference category and the first category, and determining the target category system as the second category system.

[0028] As an embodiment of the present application, the product similarity between the initial goods and the goods in the second category is obtained;

[0029] Determining the recommended goods corresponding to the initial goods from the goods in the second category according to the product similarity.

[0030] In a second aspect, the present application provides a cross-category goods recommendation device, comprising:

[0031] A processing module is configured to process training goods by using a first classification model and a second classification model, to obtain first training categories of the training goods in a first category system and second training categories of the training goods in a second category system;

[0032] A construction module is configured to determine a category association table between the first category system and the second category system according to the first training categories and the second training categories of a plurality of training goods;

[0033] The query module is configured to query the category association table according to a first category of the initial product under a first category system, to obtain a second category associated with the first category in the second category system.

[0034] The recommendation module is configured to determine a recommended product corresponding to the initial product from the products included in the second category.

[0035] In a third aspect, the present application provides a computer device, which comprises:

[0036] one or more processors;

[0037] a memory; and

[0038] one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor to implement the cross-category product recommendation method provided above.

[0039] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is loaded by a processor to execute the cross-category product recommendation method provided above.

[0040] The cross-category product recommendation method provided by the embodiments of the present application processes training products through a first classification model and a second classification model, obtains a first training category of the training products in a first category system and a second training category of the training products in a second category system, and then determines a category association table between the first category system and the second category system according to the first training category and the second training category of the plurality of training products, so that the category associated with the initial product in the second category system can be obtained according to the category association table, and the recommended product is further determined. The present application obtains the category association table by using the category classification results of the products in different category systems, so that the cross-category product recommendation can be realized, and the effect of the cross-category product recommendation is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0042] Figure 1 A step flowchart of a cross-category product recommendation method provided by the embodiments of the present application;

[0043] Figure 2 A step flowchart of a method for determining a category association table provided by the embodiments of the present application;

[0044] Figure 3 A step flow diagram for determining initial correlation degrees in an initial category association table provided by an embodiment of the present application;

[0045] Figure 4 A step flow diagram for updating initial correlation degrees in an initial category association table based on quantity proportion and proportion threshold provided by an embodiment of the present application;

[0046] Figure 5 A step flow diagram for updating a category association table in combination with external knowledge information provided by an embodiment of the present application;

[0047] Figure 6 A step flow diagram for determining a second category system provided by an embodiment of the present application;

[0048] Figure 7 A structural diagram of a cross-category commodity recommendation device provided by an embodiment of the present application;

[0049] Figure 8 A structural diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the scope of protection of the present application.

[0051] In the description of the present application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0052] In the description of the application, the word "for example" is used to mean "serving as an example, instance, or illustration." Any embodiment described as "for example" in this application is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to practice the application as claimed. In the following description, for purposes of explanation, specific details are set forth to provide a thorough understanding of the application. It will be apparent to one skilled in the art, however, that the application can be practiced without using these specific details. In other instances, well-known structures and processes are not elaborated in order not to obscure the description of the application with unnecessary details. Thus, the present application is not intended to be limited by the embodiments shown, but is to be accorded with the widest scope consistent with the principles and features disclosed herein.

[0053] To facilitate understanding of the cross-category goods recommendation method provided by the embodiments of the present application, the application scenario of the cross-category goods recommendation method is first described. Specifically, the cross-category goods recommendation is mainly applied to a goods sales platform, for example, a commonly used e-commerce platform. The goods recommendation refers to, after obtaining the consent of a user, when the user selects a certain goods to add to a shopping cart or make a purchase, recommending other goods that the user may be interested in or may need to purchase to the user according to the historical operation behavior of the user on the e-commerce platform, improving the user experience and improving the sales rate of the goods in the platform. At present, most of the algorithms for recommending goods are implemented under the same category system.

[0054] However, it is found in actual application that there are many manufacturers providing goods sales services in the platform, and different manufacturers will establish different category systems based on their own business, and the same goods will be divided into different categories with different names under different category systems, that is, the names of the categories with the same meaning under different category systems may be inconsistent, which leads to that when the goods recommendation across the category systems is needed, the categories under other category systems cannot be accurately matched according to the text or semantics, for example, when the user purchases sports pants from a clothing manufacturer on the platform, it is difficult to recommend related goods of an outdoor sports manufacturer to the user across the category systems.

[0055] To address the aforementioned issues and unify the implicit mapping relationships that may exist between various category systems within a unified platform, thereby facilitating cross-category product recommendations and improving user experience, this application provides a cross-category product recommendation method, apparatus, device, and readable storage medium. By utilizing the information carried by the product as an intermediate carrier, it indirectly finds the correspondence between categories under different category systems and generates a category association table. When cross-category product recommendations are needed, compared to semantic matching, this category association table can more accurately find the second category under other category systems that is most related to the first category to which the product to be recommended belongs, thus ensuring the accuracy of subsequent product recommendations. Specifically, the cross-category product recommendation method is typically set up as a computer program in the cross-category product recommendation apparatus. The cross-category product recommendation apparatus is typically set up as a processor in a computer device (such as the server-side device of an e-commerce platform). The product recommendation device in the computer device executes the computer program corresponding to the product recommendation method to execute the cross-category product recommendation method provided in this application embodiment.

[0056] For details, please refer to Figure 1 , Figure 1 This application provides a flowchart illustrating the steps of a cross-category product recommendation method, specifically including steps S110 to S140:

[0057] S110, the training products are processed by the first classification model and the second classification model to obtain the first training category of the training products in the first category system and the second training category of the second category system.

[0058] In this embodiment, the training products refer to a collection of various training products gathered from the platform for training and generating a category association table. Each training product typically contains multiple products, including product information such as product title, product image, product brand, and product attributes. Therefore, the more training products there are, the more product information they contain, and the more accurate the resulting category association table will be. Specifically, as a feasible implementation of this application, the product training set used in this embodiment typically contains tens of millions of products.

[0059] Considering that the step of generating the category association table provided in this application embodiment can usually be executed in advance on the platform server, that is, after the category association table is generated, the category association table can be directly stored and recorded for subsequent cross-category product recommendations.

[0060] In this embodiment, the first classification model typically refers to a pre-trained classification model that can classify goods into a specific category within a first category system. That is, the output of the first classification model corresponds to multiple categories within the first category system. The second classification model, on the other hand, refers to a pre-trained classification model that can classify goods into a specific category within a second category system. That is, the output of the second classification model corresponds to multiple categories within the second category system. Therefore, by processing the training goods using the first and second classification models respectively, the training categories of the training goods under different category systems can be obtained, such as the first training category of the first category system, the second training category of the second category system, and so on. The model structure of the classification model can adopt a common neural network model structure. This embodiment does not elaborate on the implementation scheme of the pre-trained classification model.

[0061] Of course, when it is necessary to generate a category association table between the first category system and the third category system, it is then necessary to further obtain the third classification model corresponding to the third category system. The processing of training products by the third classification model is similar to that of the first and second classification models; specifically, the training products are processed to obtain the third training category under the third category system. Since the category association tables between different category systems can all be obtained through the methods provided in the embodiments of this application, the embodiments of this application will not elaborate on other category systems, but will only use the first and second category systems as examples for explanation.

[0062] S120, determine a category association table between the first category system and the second category system based on the first training category and the second training category of the plurality of training products.

[0063] In this embodiment, after obtaining the classification results of training products under different category systems (i.e., the first training category and the second training category) using different classification models, considering that there may be multiple training products, the statistical information based on the classification results of these products can further determine the category association table between the first and second category systems. This category association table contains the association relationship between each category in the first category system and each category in the second category system. Specifically, it can be understood that the more products a category in the first category system shares with a category in the second category system, the higher the association between the two categories. For details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 This embodiment provides a flowchart illustrating the steps for determining a category association table, specifically including steps S210 to S230:

[0064] S210, Construct an initial category association table, wherein the initial category association table includes the initial association degree between each category in the first category system and each category in the second category system.

[0065] In this embodiment, the initial category association table includes the initial association degree between each category in the first category system and each category in the second category system. This initial association degree is typically a preset initial value, for example, it can be set to 0 for subsequent updates. Alternatively, as an optional implementation, the initial association degree between each category in the initial category association table can also be determined based on the semantic similarity between different categories. For details, please refer to... Figure 3 , Figure 3 This application provides a flowchart illustrating the steps for determining the initial association degree in an initial category association table, specifically including steps S310 to S320:

[0066] S310, The training categories under the first category system and the second category system are processed according to the semantic recognition model to obtain the category representation feature information corresponding to each training category.

[0067] Considering that although the names of categories with the same meaning are inconsistent in different category systems, making it impossible to obtain highly accurate similar categories directly through text or semantic matching, the category names in different category systems can still reflect the correlation between categories to some extent. Therefore, the semantic representation features of different categories can also be used to construct an initial association table. Specifically, the semantic recognition model in NLP (Natural Language Processing) technology can be used to process different categories in the first and second category systems respectively, thereby obtaining category representation feature information describing the semantic text features of each category. The category representation feature information is usually in the form of vectors.

[0068] S320, an initial association table between the first category system and the second category system is generated based on the similarity between the category representation feature information corresponding to each of the training categories.

[0069] In this embodiment, based on the aforementioned category representation feature information corresponding to the training categories, the similarity of different training categories in semantic text can be obtained by calculating the cosine angle or Euclidean distance between the category representation feature information. Simultaneously, based on this similarity, an initial association table between the first category system and the second category system can be further determined. It can be understood that when the semantic feature information of two categories in the first and second category systems is relatively similar, the initial association degree between these two categories in the initial association table can be set higher. Conversely, when the semantic feature information of two categories in the first and second category systems differs significantly, the initial association degree between these two categories in the initial association table can be set lower. This allows for further integration of semantic feature information of different categories in the first and second category systems during the subsequent generation of the category association table, thereby improving the accuracy of the association degree between the two categories.

[0070] S220, for the first target category in the first category system and the second target category in the second category system, determine the target training product from the plurality of training products.

[0071] In this embodiment, as described above, the initial category association table includes the initial association degree between each category in the first category system and each category in the second category system. Therefore, determining the category association table can also be understood as updating the initial association degree between each category in the first category system and each category in the second category system. For ease of description, this embodiment uses the first target category in the first category system and the second target category in the second category system as examples. For the first and second target categories, it is necessary to first determine the target training products from multiple training products. The first training category in the first category system is the first target category, and the second training category in the second category system is the second target category. In other words, target products classified into corresponding target categories by the corresponding category system are selected. Through this method, the target training products contained in each category in the first category system and each category in the second category system can be found.

[0072] S230, update the initial association degree in the initial category association table according to the proportion of the target training products in the training products, and obtain the category association table.

[0073] In this embodiment of the application, as described above, after determining the target training products contained in each category in the first category system and each category in the second category system, the initial correlation degree between each category in the first category system and each category in the second category system in the initial category association table is updated according to the proportion of the number of target training products, so as to obtain the final category association table used for cross-category recommendation.

[0074] It's understandable that the proportion of target training products in the training products can be interpreted as an implicit relationship between a category in the first category system and a category in the second category system. A higher proportion indicates a greater number of shared products between the two categories, thus strengthening the relationship. In this case, the initial association degree between corresponding categories in the initial category association table can be appropriately increased. Conversely, a lower proportion indicates a smaller number of shared products between the two categories, weakening the relationship. In this case, the initial association degree between corresponding categories in the initial category association table can be appropriately decreased.

[0075] Specifically, as a feasible embodiment of this application, in order to improve the accuracy of the generated category association table, the data can be properly cleaned and preprocessed. Specifically, the data can be cleaned using a certain threshold to avoid errors caused by noise. Specifically, updating the initial association degree in the initial category association table based on the proportion of target training products in the training products can also be achieved by comparing the preset proportion of target training products in the training products with a proportion threshold. That is, updating the initial association degree in the initial category association table based on the proportion of target training products in the training products to obtain the category association table includes:

[0076] Based on the comparison between the proportion of target training products in the training products and the preset proportion threshold, the initial correlation degree in the initial category association table is updated to obtain the category association table.

[0077] Specifically, based on the comparison between the proportion of target training products in the training products and the preset proportion threshold, the initial correlation degree in the initial category association table is updated to obtain the category association table.

[0078] For details, please refer to Figure 4 , Figure 4 This application provides a flowchart illustrating the steps for updating the initial association degree in the initial category association table based on quantity proportion and ratio threshold, specifically including steps S410 to S420:

[0079] S410, if the proportion of the target training product in the training products is greater than a preset proportion threshold, the initial correlation degree in the initial category association table is updated according to the proportion of the number of products to obtain the category association table.

[0080] In this embodiment of the application, the proportion threshold can be used to clean data that is below the threshold. That is, only when the proportion of the number of target training products in the training products is greater than the preset proportion threshold, i.e. when the amount of data is large enough, will the initial correlation degree in the initial category association table be updated according to the proportion of the number to obtain the category association table. For example, the value of the proportion of the number is added to the initial correlation degree, or the initial correlation degree is updated to the value of the proportion of the number.

[0081] S420, if the proportion of the target training product in the training products is less than or equal to a preset proportion threshold, the initial correlation degree in the initial category association table is updated to the preset correlation degree.

[0082] In this embodiment of the application, conversely, when the proportion of the number of target training products in the training products is less than or equal to the preset proportion threshold, the correlation between the category under the first category system and the category under the second category system corresponding to the classification result is low. This data can be regarded as abnormal data. Therefore, it is possible to directly select to update the initial correlation in the initial category correlation table to the preset correlation, such as updating it to 0, directly keeping the initial correlation unchanged, etc.

[0083] Specifically, the ratio threshold can usually be a preset value, such as 0.05. Of course, in actual business, the frequency threshold can also be dynamically set based on the product information of the initial product, which will not be elaborated here in the embodiments of this application.

[0084] In this embodiment of the application, by using a proportional threshold to clean the data, the accuracy of the obtained category association table can be further improved, and it can be further used for more accurate cross-category product recommendations in the future.

[0085] Of course, based on the category association table obtained through the aforementioned methods, further refinement of the table can be achieved by incorporating external knowledge from the e-commerce field. For details, please refer to [link to relevant documentation]. Figure 5 , Figure 5 This application provides a flowchart illustrating the steps for updating a category association table by incorporating external knowledge information, specifically including steps S510 to S520:

[0086] S510, Obtain external knowledge information of the first category system and the second category system.

[0087] In this embodiment of the application, the external knowledge information of the first category system and the second category system can generally be understood as an indicator map in the e-commerce field. For example, it can usually include category product proportion information, that is, the proportion of the number of products contained in each category, and category brand proportion, that is, the proportion of the number of brand products in each category. Of course, external knowledge information can also include more, such as statistical information on category product prices or statistical information on category product attributes, etc. This embodiment of the application does not make specific restrictions on external knowledge information.

[0088] S520, the initial correlation degree in the initial category association table is updated according to the external knowledge information and the proportion of the target training products in the training products, so as to obtain the category association table.

[0089] In this embodiment of the application, after obtaining the external knowledge information, the initial correlation degree in the initial category association table is updated by combining the external knowledge information with the proportion of the target training products in the training products. This allows the category association table to learn knowledge information in the e-commerce field, thereby outputting a more accurate category association table.

[0090] Specifically, updating the initial correlation degree in the initial category association table based on the external knowledge information and the proportion of the target training products in the training products can be achieved by inputting the external knowledge information, the proportion of the products, and the initial category association table into a pre-trained ensemble model, thereby outputting the category association table.

[0091] The aforementioned implementation scheme has specifically explained the complete implementation process of generating the category association table between the first category system and the second category system. In fact, the above content can also be used as a reference for the category association table between other different category systems. The cross-category product recommendation device can pre-calculate the category association table between different category systems and store these category association tables in the database in advance, thereby establishing a mapping relationship between different category systems within the platform.

[0092] S130, based on the initial product to be recommended, query the category association table under the first category system to obtain the second category in the second category system that is associated with the first category.

[0093] In this embodiment, the initial products to be recommended typically refer to products that users may be interested in, as determined by specific user actions on the platform, such as adding to the shopping cart, making a direct purchase, or marking notes. Therefore, the platform needs to recommend similar products to users to increase user stickiness. This embodiment will not elaborate on the implementation scheme for determining the initial products to be recommended.

[0094] In this embodiment, after determining the initial product, the first category system and the first category under this first category system can be determined based on the product details, such as manufacturer brand information. For example, when a user selects men's athletic pants from a certain clothing manufacturer, it can be determined that the first category of the initial product under the first category system established by the clothing manufacturer is men's pants. At this time, when recommendations need to be made under the same category system, other products under this first category, such as casual men's pants, can be conveniently recommended to the user. However, when it is necessary to recommend products from other manufacturers to the user, since the category systems established by other manufacturers may not directly contain the category name "men's pants" or the product name "athletic men's pants," it is easy to cause product recommendations to be biased, affecting the user experience.

[0095] The solution provided in this application establishes a category association table describing the relationships between different categories in the first category system and different categories in the second category system. Therefore, by querying the category association table based on the first category, the second category associated with the first category in the second category system can be obtained. It should be noted that since there are many category systems in the e-commerce field, the second category system described here is merely illustrative. In fact, the determined second category system can also be obtained based on category association tables between multiple different category systems and the first category system. For details, please refer to [link to relevant documentation]. Figure 6 , Figure 6 This application provides a flowchart illustrating the steps for determining a second category system, specifically including steps S610 to S620:

[0096] S610, based on the initial product to be recommended, query multiple category association tables under the first category system to obtain the reference category with the highest correlation to the first category under each candidate category system.

[0097] In this embodiment of the application, as can be seen from the foregoing description, after establishing the mapping relationship between different category systems within the platform, there will be multiple category association tables associated with the first category system, such as the category association table between the second category system and the first category system, the category association table between the third category system and the first category system, and so on. At this time, by querying multiple category association tables based on the first category, the reference category with the highest correlation to the first category under each candidate category system can be determined from the different category association tables.

[0098] S620, based on the correlation between the first category system and each of the candidate category systems and the correlation between the reference category and the first category, a target category system is determined from the candidate category systems, and the target category system is determined as the second category system.

[0099] In this embodiment of the application, the correlation between the first category system and each candidate category system describes the correlation between category systems, which can usually be predicted by a neural network model. The correlation between the reference category and the first category describes the correlation between categories. Based on the weighted sum of the two, a second category system that is more similar to the first category system can be determined from the candidate category systems for subsequent cross-category product recommendations.

[0100] S140, determine the recommended product corresponding to the initial product from the products included in the second category.

[0101] In this embodiment of the application, after obtaining the second category that is closest to the first category under other category systems, the recommended product corresponding to the initial product can be determined based on the products in the second category. For example, in the simplest case, all or a random portion of the products in the second category can be recommended to the user. Of course, as another feasible embodiment of this application, more accurate product recommendations can be made based on the similarity between products. Specifically, the method includes:

[0102] Obtain the product similarity between the initial product and the products in the second category.

[0103] In this embodiment of the application, the similarity between products can be obtained by calculating the similarity between the semantic representation vectors of product information. For example, by processing product information through the aforementioned semantic recognition model, the semantic representation vector of the product can be obtained. Then, by using the cosine similarity or Euclidean distance between the semantic representation vectors of different products, the product similarity between different products can be calculated.

[0104] Based on the product similarity, the recommended product corresponding to the initial product is determined from the products in the second category.

[0105] In this embodiment of the application, based on product similarity, the product with the highest similarity to the initial product in the second category can be used as the recommended product, or all products in the second category with a similarity higher than a preset threshold to the initial product can be used as the recommended product. This embodiment of the application will not be elaborated further here.

[0106] The cross-category product recommendation method provided in this application processes training products using a first classification model and a second classification model to obtain a first training category for the training products in a first category system and a second training category in a second category system. Then, based on the first and second training categories of multiple training products, a category association table between the first and second category systems is determined. This table allows for querying the categories associated with the initial products under the second category system, thereby further determining the recommended products. This application utilizes the category classification results of products under different category systems to obtain the category association table, thus enabling cross-category product recommendation and further improving its effectiveness.

[0107] To better implement the cross-category product recommendation method provided in the embodiments of this application, this application also provides a cross-category product recommendation device, such as... Figure 7 As shown, the cross-category product recommendation device 700 includes:

[0108] The processing module 710 is used to process the training products through the first classification model and the second classification model to obtain the first training category of the training products in the first category system and the second training category in the second category system.

[0109] The construction module 720 is used to determine a category association table between the first category system and the second category system based on the first training category and the second training category of the multiple training products.

[0110] The query module 730 is used to query the category association table based on the first category under the first category system of the initial product to be recommended, and obtain the second category in the second category system that is associated with the first category;

[0111] The recommendation module 740 is used to determine recommended products corresponding to the initial products from the products included in the second category.

[0112] In some embodiments of this application, the construction module 720 is further configured to construct an initial category association table, the initial category association table including the initial association degree between each category in the first category system and each category in the second category system; for the first target category in the first category system and the second target category in the second category system, a target training product is determined from the plurality of training products; wherein, the first training category of the target training product under the first category system is the first target category, and the second training category under the second category system is the second target category; the initial association degree in the initial category association table is updated according to the proportion of the number of target training products in the training products to obtain the category association table.

[0113] In some embodiments of this application, the construction module 720 is further configured to update the initial correlation degree in the initial category association table based on the comparison result of the proportion of the number of target training products in the training products and the preset proportion threshold, so as to obtain the category association table.

[0114] In some embodiments of this application, the construction module 720 is further configured to, when the proportion of the number of target training products in the training products is greater than a preset proportion threshold, update the initial correlation degree in the initial category association table according to the proportion of the number of products to obtain the category association table; and when the proportion of the number of target training products in the training products is less than or equal to the preset proportion threshold, update the initial correlation degree in the initial category association table to a preset correlation degree.

[0115] In some embodiments of this application, the construction module 720 is further configured to process the training categories under the first category system and the second category system according to the semantic recognition model to obtain category representation feature information corresponding to each training category; and generate an initial association table between the first category system and the second category system according to the similarity between the category representation feature information corresponding to each training category.

[0116] In some embodiments of this application, the construction module 720 is further configured to obtain external knowledge information of the first category system and the second category system, wherein the external knowledge information includes at least one of category product proportion information and category brand proportion information in the first category system and the second category system; and update the initial correlation degree in the initial category association table according to the external knowledge information and the number proportion of target training products in the training products to obtain the category association table.

[0117] In some embodiments of this application, the query module 730 is further configured to query multiple category association tables based on the first category under the first category system of the initial product to be recommended, and obtain the reference category with the highest correlation with the first category under each candidate category system; determine the target category system from the candidate category system based on the correlation between the first category system and each candidate category system and the correlation between the reference category and the first category system, and determine the target category system as the second category system.

[0118] In some embodiments of this application, the recommendation module 740 is further configured to obtain the product similarity between the initial product and the products in the second category; and determine the recommended product corresponding to the initial product from the products in the second category based on the product similarity.

[0119] Specific limitations regarding the cross-category product recommendation device can be found in the limitations of the cross-category product recommendation method described above, and will not be repeated here. Each module in the aforementioned cross-category product recommendation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0120] In some embodiments of this application, the cross-category product recommendation device 700 can be implemented as a computer program, which can be implemented in, for example... Figure 8 The computer device shown operates on this device. The computer device's memory can store the various program modules that make up the cross-category product recommendation device 700, for example, Figure 7 The processing module 710, construction module 720, query module 730, and recommendation module 740 are shown. The computer program comprised of these modules causes the processor to execute the steps of the cross-category product recommendation methods described in the various embodiments of this application.

[0121] For example, Figure 8 The computer device shown can be used as follows Figure 7 The processing module 710 in the cross-category product recommendation device 700 shown executes step S110. The computer device can execute step S120 via the construction module 720. The computer device can execute step S130 via the query module 730. The computer device can execute step S140 via the recommendation module 740. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with external computer devices via a network connection. When the computer program is executed by the processor, it implements a cross-category product recommendation method.

[0122] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0123] In some embodiments of this application, a computer device is provided, including one or more processors; memory; and one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor to perform the following steps:

[0124] The training products are processed by the first classification model and the second classification model to obtain the first training category of the training products in the first category system and the second training category of the second category system.

[0125] A category association table between the first category system and the second category system is determined based on the first training category and the second training category of the multiple training products;

[0126] Based on the initial product to be recommended, query the category association table under the first category system to obtain the second category associated with the first category in the second category system;

[0127] Determine the recommended product corresponding to the initial product from the products included in the second category.

[0128] In some embodiments of this application, when the processor executes the computer program, it further performs the following steps: constructing an initial category association table, the initial category association table including the initial association degree between each category in the first category system and each category in the second category system; determining target training products from the plurality of training products for a first target category in the first category system and a second target category in the second category system; wherein, the first training category of the target training product under the first category system is the first target category, and the second training category under the second category system is the second target category; updating the initial association degree in the initial category association table according to the proportion of the number of target training products in the training products, thereby obtaining the category association table.

[0129] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: updating the initial association degree in the initial category association table based on the comparison result between the proportion of the number of target training products in the training products and the preset proportion threshold, thereby obtaining the category association table.

[0130] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: when the proportion of the number of target training products in the training products is greater than a preset proportion threshold, the initial correlation degree in the initial category association table is updated according to the proportion of the number to obtain the category association table; when the proportion of the number of target training products in the training products is less than or equal to the preset proportion threshold, the initial correlation degree in the initial category association table is updated to a preset correlation degree.

[0131] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: processing the training categories under the first category system and the second category system according to the semantic recognition model to obtain category representation feature information corresponding to each training category; generating an initial association table between the first category system and the second category system based on the similarity between the category representation feature information corresponding to each training category.

[0132] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: obtaining external knowledge information of the first category system and the second category system, wherein the external knowledge information includes at least one of category product proportion information and category brand proportion information in the first category system and the second category system; updating the initial correlation degree in the initial category association table according to the external knowledge information and the quantity proportion of target training products in the training products, thereby obtaining the category association table.

[0133] In some embodiments of this application, when the processor executes the computer program, it further performs the following steps: querying multiple category association tables based on the first category under the first category system of the initial product to be recommended, and obtaining the reference category with the highest correlation with the first category under each candidate category system; determining the target category system from the candidate category systems based on the correlation between the first category system and each candidate category system and the correlation between the reference category and the first category system, and determining the target category system as the second category system.

[0134] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: obtaining the product similarity between the initial product and the products in the second category; and determining the recommended product corresponding to the initial product from the products in the second category based on the product similarity.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0136] In some embodiments of this application, a computer-readable storage medium is provided, storing a computer program that is loaded by a processor, causing the processor to perform the following steps:

[0137] The training products are processed by the first classification model and the second classification model to obtain the first training category of the training products in the first category system and the second training category of the second category system.

[0138] A category association table between the first category system and the second category system is determined based on the first training category and the second training category of the multiple training products;

[0139] Based on the initial product to be recommended, query the category association table under the first category system to obtain the second category associated with the first category in the second category system;

[0140] Determine the recommended product corresponding to the initial product from the products included in the second category.

[0141] In some embodiments of this application, when the processor executes the computer program, it further performs the following steps: constructing an initial category association table, the initial category association table including the initial association degree between each category in the first category system and each category in the second category system; determining target training products from the plurality of training products for a first target category in the first category system and a second target category in the second category system; wherein, the first training category of the target training product under the first category system is the first target category, and the second training category under the second category system is the second target category; updating the initial association degree in the initial category association table according to the proportion of the number of target training products in the training products, thereby obtaining the category association table.

[0142] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: updating the initial association degree in the initial category association table based on the comparison result between the proportion of the number of target training products in the training products and the preset proportion threshold, thereby obtaining the category association table.

[0143] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: when the proportion of the number of target training products in the training products is greater than a preset proportion threshold, the initial correlation degree in the initial category association table is updated according to the proportion of the number to obtain the category association table; when the proportion of the number of target training products in the training products is less than or equal to the preset proportion threshold, the initial correlation degree in the initial category association table is updated to a preset correlation degree.

[0144] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: processing the training categories under the first category system and the second category system according to the semantic recognition model to obtain category representation feature information corresponding to each training category; generating an initial association table between the first category system and the second category system based on the similarity between the category representation feature information corresponding to each training category.

[0145] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: obtaining external knowledge information of the first category system and the second category system, wherein the external knowledge information includes at least one of category product proportion information and category brand proportion information in the first category system and the second category system; updating the initial correlation degree in the initial category association table according to the external knowledge information and the quantity proportion of target training products in the training products, thereby obtaining the category association table.

[0146] In some embodiments of this application, when the processor executes the computer program, it further performs the following steps: querying multiple category association tables based on the first category under the first category system of the initial product to be recommended, and obtaining the reference category with the highest correlation with the first category under each candidate category system; determining the target category system from the candidate category systems based on the correlation between the first category system and each candidate category system and the correlation between the reference category and the first category system, and determining the target category system as the second category system.

[0147] In some embodiments of this application, when the processor executes the computer program, it further implements the following steps: obtaining the product similarity between the initial product and the products in the second category; and determining the recommended product corresponding to the initial product from the products in the second category based on the product similarity.

[0148] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The foregoing has provided a detailed description of a cross-category product recommendation method, apparatus, computer device, and storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A cross-category product recommendation method, characterized in that, The method comprises: processing the training commodities through the first classification model and the second classification model to obtain first training categories of the training commodities in the first category system and second training categories of the training commodities in the second category system; determining a category association table between the first category system and the second category system according to the first training categories and the second training categories of the plurality of training commodities; querying the category association table according to a first category of an initial commodity to be recommended in the first category system to obtain a second category associated with the first category in the second category system; determining a recommended commodity corresponding to the initial commodity from commodities included in the second category.

2. The product recommendation method according to claim 1, characterized by, The method comprises: constructing an initial category association table, wherein the initial category association table comprises initial association degrees between categories in the first category system and categories in the second category system; determining target training commodities from the plurality of training commodities for a first target category in the first category system and a second target category in the second category system, wherein the first training category of the target training commodities in the first category system is the first target category, and the second training category of the target training commodities in the second category system is the second target category; updating the initial association degrees in the initial category association table according to a proportion of the target training commodities in the training commodities to obtain the category association table.

3. The product recommendation method according to claim 2, characterized by, The method comprises: updating the initial association degrees in the initial category association table according to a comparison result of the proportion of the target training commodities in the training commodities and a preset proportion threshold to obtain the category association table.

4. The product recommendation method according to claim 3, characterized by, The method comprises: in a case where the proportion of the target training commodities in the training commodities is greater than the preset proportion threshold, updating the initial association degrees in the initial category association table according to the proportion to obtain the category association table; in a case where the proportion of the target training commodities in the training commodities is less than or equal to the preset proportion threshold, updating the initial association degrees in the initial category association table to a preset association degree.

5. The product recommendation method according to claim 2, characterized by, The method comprises: processing the training categories in the first category system and the second category system according to a semantic recognition model to obtain category representation feature information corresponding to each training category; generating an initial association table between the first category system and the second category system according to similarities between the category representation feature information corresponding to each training category.

6. The commodity recommendation method according to claim 2, characterized by, The method comprises: obtaining external knowledge information of the first category system and the second category system, the external knowledge information including at least one of category commodity proportion information and category brand proportion information in the first category system and the second category system; updating initial correlation degrees in the initial category correlation table according to the external knowledge information and a quantity proportion of a target training commodity in the training commodities, to obtain the category correlation table.

7. The product recommendation method according to claim 1, characterized by, Before the method further includes querying the category correlation table according to a first category of the initial commodity in the first category system, to obtain a second category associated with the first category in the second category system. Before the method further includes querying a plurality of category correlation tables according to a first category of the initial commodity in the first category system, to respectively obtain a reference category with a highest correlation degree with the first category in each candidate category system. According to correlation degrees between the first category system and each candidate category system and a correlation degree between the reference category and the first category, a target category system is determined from the candidate category systems, and the target category system is determined as the second category system.

8. The product recommendation device according to any one of claims 1 to 7, characterized in that The method further includes: obtaining a commodity similarity between the initial commodity and commodities in the second category; determining a recommended commodity corresponding to the initial commodity from the commodities in the second category according to the commodity similarity. 9.A cross-category item recommendation apparatus, characterized by comprising: The method further includes: a processing module configured to process training commodities through a first classification model and a second classification model, to obtain first training categories of the training commodities in a first category system and second training categories of the training commodities in a second category system; a construction module configured to determine a category correlation table between the first category system and the second category system according to the first training categories and the second training categories of a plurality of the training commodities; a querying module configured to query the category correlation table according to a first category of an initial commodity to be recommended in the first category system, to obtain a second category associated with the first category in the second category system; a recommendation module configured to determine a recommended commodity corresponding to the initial commodity from commodities included in the second category.

10. A computer device, comprising: The computer device includes: one or more processors; a memory; and one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor to implement the cross-category commodity recommendation method in any one of claims 1 to 8.

11. A computer readable storage medium characterized by, A computer program is stored thereon, and the computer program is loaded by a processor to execute the cross-category commodity recommendation method in any one of claims 1 to 8.

Images