Article size recommendation method, article size table generation method, and article size table generation device

Abstract

This specification provides a product size recommendation method, a product size table generation method, and an apparatus. The product size recommendation method includes: in response to a size viewing event of a target product, displaying a first size table, a table conversion control, and a size recommendation information block for the target product. The first size table is obtained by inputting a first target prompt word into a target generation model. The first target prompt word is constructed based on original size information and a first preset prompt word template. The size recommendation information block displays recommended sizes for the current account, determined based on the account information of the current account and the first size table. In response to a trigger operation on the table conversion control, switching the first size table to a second size table. The second size table is a table after switching the language and / or size units of the first size table. This provides a clear and accurate size table and recommended sizes, and offers language and / or size unit switching functionality, improving the accuracy and efficiency of size selection.

Description

Technical Field

[0001] This specification relates to the field of computer technology, and in particular to a product size recommendation method, a product size table generation method and apparatus. Background Technology

[0002] With the continuous development of computer and internet technology, more and more users can trade various goods related to clothing, food, housing, and transportation online. E-commerce platforms deal with footwear and apparel products, which involve product sizes. Different merchants use their own methods (such as pictures, text, different units, etc.) to provide different styles of size charts or size recommendations. Users select footwear and apparel products based on the displayed size charts or size recommendations.

[0003] In existing technologies, the product size information provided by merchants is presented in complex and diverse formats, with chaotic layouts and redundant information, failing to provide users with clear and accurate size references. Users have to go through cumbersome operations to select sizes, which is inconvenient for users and results in low accuracy and efficiency in size selection. Therefore, there is an urgent need for a more accurate and efficient product size recommendation solution. Summary of the Invention

[0004] In view of this, embodiments of this specification provide a product size recommendation method. One or more embodiments of this specification also relate to a product size chart generation method, a product size recommendation device, a product size chart generation device, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0005] According to a first aspect of the embodiments of this specification, a product size recommendation method is provided, including:

[0006] In response to a size viewing event of a target product, a first size table, a table conversion control, and a size recommendation information block of the target product are displayed. The first size table is obtained by inputting a first target prompt word into a target generation model. The first target prompt word is constructed based on the original size information and a first preset prompt word template. The size recommendation information block displays recommended sizes for the current account. The recommended sizes are determined based on the account information of the current account and the first size table.

[0007] In response to a trigger operation on the table conversion control, the first size table is switched to a second size table; wherein the second size table is a table after the language and / or size units of the first size table have been switched.

[0008] According to a second aspect of the embodiments of this specification, a method for generating a product size chart is provided, comprising:

[0009] Obtain the original size information of the target product, wherein the original size information is an original size image and / or an original size description text;

[0010] The original size information of the target product and the second target prompt word are input into the target extraction model to obtain the structured size data of the original size information, and the first target prompt word is constructed based on the structured size data of the original size information and the first preset prompt word template.

[0011] The first target prompt word is input into the target generation model to obtain the first size table of the target product, wherein the target generation model is trained by at least two styles of sample size information.

[0012] According to a third aspect of the embodiments of this specification, a product size recommendation device is provided, comprising:

[0013] The first display module is configured to respond to a size viewing event of a target product by displaying a first size table, a table conversion control, and a size recommendation information block for the target product. The first size table is obtained by inputting a first target prompt word into a target generation model. The first target prompt word is constructed based on the original size information and a first preset prompt word template. The size recommendation information block displays recommended sizes for the current account. The recommended sizes are determined based on the account information of the current account and the first size table.

[0014] The switching module is configured to switch the first size table to a second size table in response to a trigger operation of the table conversion control; wherein the second size table is a table after switching the language and / or size units of the first size table.

[0015] According to a fourth aspect of the embodiments of this specification, an apparatus for generating a product size chart is provided, comprising:

[0016] The acquisition module is configured to acquire the original size information of the target product, wherein the original size information is an original size image and / or original size description text.

[0017] The construction module is configured to input the original size information of the target product and the second target prompt word into the target extraction model to obtain the structured size data of the original size information, and to construct the first target prompt word based on the structured size data of the original size information and the first preset prompt word template;

[0018] The generation module is configured to input the first target prompt word into the target generation model to obtain a first size table of the target product, wherein the target generation model is trained using sample size information of at least two styles.

[0019] According to a fifth aspect of the embodiments of this specification, a computing device is provided, comprising:

[0020] Memory and processor;

[0021] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the above-described product size recommendation method or product size table generation method.

[0022] According to a sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-described product size recommendation method or product size table generation method.

[0023] According to a seventh aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described product size recommendation method or product size table generation method.

[0024] This specification provides a product size recommendation method. In response to a size viewing event of a target product, it displays a first size table, a table conversion control, and a size recommendation information block for the target product. The first size table is obtained by inputting a first target prompt word into a target generation model. The first target prompt word is constructed based on the original size information and a first preset prompt word template. The size recommendation information block displays recommended sizes for the current account, determined based on the account information of the current account and the first size table. In response to a trigger operation on the table conversion control, the first size table is switched to a second size table. The second size table is a table after switching the language and / or size units of the first size table.

[0025] One embodiment of this specification implements a method to construct a first target prompt word based on the original size information of the target product and a first preset prompt word template. Inputting the first target prompt word into a target generation model yields a first size table for the target product. This unifies the original size information, which is often presented in complex and varied formats, has disorganized layouts, and is informationally redundant, into a first size table with a corresponding structure. This provides users with a clear and accurate size table. Furthermore, based on the current account's information and the first size table, the recommended size for the current account is determined and displayed in the size recommendation information block, thus providing accurate size recommendations without cumbersome operations. Additionally, a table conversion control can be displayed, providing the ability to switch the language and / or size units of the first size table, offering a size table that better suits the user's actual language and / or size habits, greatly improving the accuracy and efficiency of size selection. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a product size recommendation method provided in one embodiment of this specification;

[0027] Figure 2 This is a schematic diagram of original size information provided in one embodiment of this specification;

[0028] Figure 3 This is a schematic diagram illustrating the generation process of a first size chart for a target product according to one embodiment of this specification;

[0029] Figure 4 This is a schematic diagram of a product details page for a target product provided in one embodiment of this specification;

[0030] Figure 5 This is a schematic diagram of a size chart in a different language version provided in one embodiment of this specification;

[0031] Figure 6 This is a schematic diagram illustrating the switching of the media content display area in a product details page, provided in one embodiment of this specification.

[0032] Figure 7 This is a schematic diagram illustrating the switching of the media content display area in another product details page, provided in one embodiment of this specification.

[0033] Figure 8 This is a schematic diagram illustrating the switching of a size chart according to one embodiment of this specification;

[0034] Figure 9 This is a flowchart illustrating a method for generating a product size chart according to one embodiment of this specification;

[0035] Figure 10 This is a flowchart of a product size recommendation method provided in one embodiment of this specification;

[0036] Figure 11 This is a schematic diagram of the structure of a product size recommendation device provided in one embodiment of this specification;

[0037] Figure 12 This is a schematic diagram of a device for generating a product size chart according to one embodiment of this specification;

[0038] Figure 13 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0039] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0040] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0041] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0042] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0043] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0044] "In response to": This indicates the state where a corresponding event occurs or a condition is met. The timing of subsequent actions performed in response to this event or condition is not necessarily strongly correlated with the time when the event occurs or the condition is met. For example, in some cases, subsequent actions may be performed immediately when the event occurs or the condition is met; while in other cases, subsequent actions may be performed some time after the event occurs or the condition is met.

[0045] Triggered operations refer to actions performed on information (such as controls) provided by a computer device. These operations send corresponding instructions to the computer device, triggering it to execute the next task. The next task triggered by different triggered operations on different information can all be pre-programmed. These triggered operations can be performed manually by the user; for example, clicks, double-clicks, long presses, and swipes performed by the user on content displayed on the computer screen are all triggered operations. In some cases, these triggered operations can also be executed by the computer device based on a pre-defined program.

[0046] An event is a specific action or change that occurs in a system or application. It can be triggered by the user (such as clicking a button or typing on the keyboard) or by the system (such as a timer expiring or a file download completing). Programs typically listen for these events and execute corresponding code when a specific event occurs.

[0047] Controls: In computer user interfaces, a "control" refers to any element that a user can directly interact with. These elements allow users to input data, select options, or trigger certain actions. Controls can be graphical (such as buttons, text boxes, checkboxes, etc.) or more abstract (such as voice commands). Buttons are typically used to perform an operation or open a new interface; text boxes allow users to input or edit text information; checkboxes represent a Boolean value (yes / no), which users can select or deselect; dropdown lists provide a list of options for users to choose from, usually displaying only one selected value; list boxes display multiple options, allowing users to select one or more items; labels display static text, usually used to explain the function of other controls, and can also be used as buttons; scrollbars allow users to navigate large datasets, such as browsing long documents or lists; images display static pictures and can sometimes be used as buttons; combo boxes combine the functionality of text boxes and dropdown lists, allowing users to both input text and select from a list. These controls are typically provided with standard styles by the operating system or development kit. Developers can customize their appearance and behavior as needed. Different operating systems and programming languages ​​(such as Java, C#, Python, etc.) have their own set of controls and provide corresponding APIs to create and manage these controls.

[0048] Content Application Platform: This is an application platform that integrates various types of content. The content application platform provides users with a rich variety of multimedia content, including but not limited to live streaming, on-demand video, audio, text and image information, and social interaction.

[0049] Multimedia content refers to content that includes multiple media formats, including but not limited to text, images, audio, and video. Multimedia technology integrates these different media types to provide a richer and more interactive user experience.

[0050] Large Models: Large-scale pre-trained natural language models based on deep learning, with powerful contextual understanding and generation capabilities, capable of handling complex and diverse natural language tasks under the guidance of prompts.

[0051] Multimodal large-scale models: Based on deep learning technology, these are large-scale models that integrate the processing capabilities of multimodal data such as images and text. They can process and integrate various data modalities, such as text, images, audio, and video. Compared with traditional single-modal models, multimodal large-scale models can utilize complementary information between different modalities to achieve more complex and intelligent tasks. Multimodal large-scale models are typically trained on large-scale datasets to learn the features and relationships between different modalities. In the embodiments of this specification, a multimodal large-scale model is used to accurately identify and parse unstructured size information and generate corresponding structured size information.

[0052] Structured size chart: A structured size chart refers to a size data table with a fixed format and logical relationship. Through multimodal large model, unstructured size information (such as images and text) uploaded by merchants can be parsed, and a standardized, fixed-format, logically related, and applicable structured size chart is generated after automated data cleaning.

[0053] User size profiles refer to a file specifically created for each user, recording their various size information; in other words, a personal size data set. This data set may cover clothing sizes (such as height, weight, chest, waist, and shoulder width of tops, and inseam and hip circumference of trousers), shoe sizes, and may also include sizes suitable for specific items. Creating such profiles helps in providing accurate size recommendations to users and also makes it easier for users to understand and manage their own size data.

[0054] Size Assistant: This is an auxiliary tool that is usually used to help users determine the appropriate product size. It is an interactive module that can provide personalized size recommendations based on user size profiles and product size data through models and adaptive learning mechanisms.

[0055] It should be noted that with the rapid development of footwear and apparel categories on e-commerce platforms, the presentation of size information on product detail pages has become increasingly complex and diverse, encompassing various non-standardized images uploaded by merchants and custom text descriptions. However, due to the lack of unified standards, this information generally suffers from problems such as unstructured formatting, chaotic layout, and redundant information. Product detail pages lack intelligent parsing and unified integration of size data, failing to generate intuitive, structured size charts. This results in a lack of clear and accurate size references for users, requiring them to repeatedly compare images, manually convert units, or consult customer service to make a preliminary assessment of fit, leading to inefficient decision-making and frequent returns and exchanges due to unsuitable sizes.

[0056] In one implementation, merchants may not fill in size data in the product backend according to specifications, but instead directly provide size images. The size images provided by different merchants are complex and diverse in presentation, with chaotic layout and redundant information. OCR (Optical Character Recognition) technology can be used to analyze, extract features, and match patterns in the characters in the size images provided by merchants to identify the size content and convert it into a text format that computers can process. Traditional OCR technology suffers from low recognition efficiency and poor accuracy due to insufficient single-modal parsing capabilities. OCR requires separate training for each different table type. If the table presentation in the size image is slightly different, the required size data cannot be extracted, making it difficult to achieve accurate size recommendations.

[0057] In the embodiments described in this specification, a multimodal large model can be used to fuse multi-source data such as images and text, achieving efficient parsing and accurate structured output of unstructured size information. This allows for efficient parsing and intelligent classification of unstructured size information, dynamically presenting structured size data in the core display area of ​​the product details page. Furthermore, a dynamic user size profile can be built, combining product parameters and user behavior data to generate high-precision matching recommendations, reducing the risk of decision-making bias. This achieves end-to-end optimization from "data parsing - user matching - accurate recommendation," a technological upgrade from "static matching" to "dynamic adaptation," significantly improving the efficiency and accuracy of user size decisions, enhancing users' shopping decision-making efficiency, and helping users find the most suitable size faster and more accurately. In addition, a structured size table is embedded in the core area of ​​the product details page, supporting unit switching, multi-language adaptation, and intelligent recommendation modules. A modular interactive framework integrates the structured size table, user profile management, and size recommendation functions, using visualized data presentation and real-time recommendation mechanisms to enhance user trust and decision-making efficiency, facilitating users to quickly select suitable sizes and assisting in user decision-making. In this way, users can quickly find the right size through clear, structured data and accurate recommendations, reducing repeated comparisons and accidental purchases, significantly lowering user decision-making costs and return risks. Users can verify the recommendation logic (such as viewing the algorithm basis and similar buyer cases), reducing decision-making anxiety caused by information asymmetry, enhancing platform transparency and user trust. This increases user confidence in product suitability, promotes purchase conversion and repurchase rates, while also driving the localization and adaptation capabilities of cross-border goods, strengthening platform competitiveness, and optimizing the platform's shopping experience and business ecosystem.

[0058] This specification provides a product size recommendation method, and also relates to a product size chart generation method, a product size recommendation device, a product size chart generation device, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0059] See Figure 1 , Figure 1 A flowchart of a product size recommendation method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0060] Step 102: In response to the size viewing event of the target product, display the first size table, table conversion control and size recommendation information block of the target product. The first size table is obtained by inputting the first target prompt word into the target generation model. The first target prompt word is constructed based on the original size information and the first preset prompt word template. The size recommendation information block displays the recommended size for the current account. The recommended size is determined based on the account information of the current account and the first size table.

[0061] Specifically, the target product is any product on the content application platform, such as clothing, shoes, or specific items available in different sizes. The target product size viewing event refers to the event triggered by actions related to viewing the size of the target product. This event can include at least one of the following: triggering events of the size control on the target product's product selection page, triggering operations of the size control on the target product's product details page, or switching operations of the media content display area on the target product's product details page. The original size information can be an original size image and / or original size description text.

[0062] It should be noted that the original size information provided by merchants is presented in a complex and varied manner, with chaotic layout and redundant information. (Example provided). Figure 2 This is a schematic diagram of original size information provided in one embodiment of this specification. The original size information provided by the merchant is an original size image, such as... Figure 2 As shown, the size charts in the nine original size images are all different, which cannot provide users with a clear and accurate reference for product sizes.

[0063] In practice, a target generation model, pre-trained with sample size information from at least two different styles, can be used to construct a first target prompt word based on the original size information and a first preset prompt word template. This first target prompt word is then input into the target generation model to obtain a first size table, thus converting unstructured original size information into a structured first size table. The target generation model is a large-scale model capable of accurately identifying and parsing the original size information in the first target prompt word using deep learning technology, and generating the corresponding structured first size table according to the instructions of the first target prompt word. The first target prompt word refers to the prompt information from the target generation model, instructing it to process the original size information according to the structure of the size table to be generated, thus producing the corresponding structured first size table. The first preset prompt word template is a prompt word template pre-configured based on the structure of the size table to be generated. This template reserves insertion positions for the original size information, allowing the structured size data from the original size information to be inserted into these positions to obtain the first target prompt word.

[0064] In one optional implementation of this embodiment, the first target prompt word is constructed based on the original size information and the first preset prompt word template, including: the first target prompt word is constructed based on the structured size data of the original size information and the first preset prompt word template, the structured size data of the original size information is obtained by inputting the original size information and the second target prompt word into the target extraction model, and the second target prompt word is constructed based on the extraction method of the structured size data in the original size information.

[0065] In practice, a second target prompt word can be constructed based on the extraction method of structured size data from the original size information and a second preset prompt word template. The original size information and the second target prompt word are then input into the target extraction model to obtain the structured size data from the original size information. This target extraction model is pre-trained and integrates the processing capabilities of multimodal original size information such as images and text, enabling it to accurately identify and parse unstructured size information from the original size information and extract structured size data. This structured size data can then be filled into the first preset prompt word template to construct the second target prompt word. Inputting the second target prompt word into the target generation model yields the first size table for the target product.

[0066] The second target prompt word serves as a prompt for the target extraction model, instructing it to process the input raw size information according to the extraction method of structured size data in the original size information, and obtain structured size data that conforms to the table features. The first target prompt word includes the structured size data extracted by the target extraction model, instructing the target generation model to process the structured size data according to the structural information of the size table to be generated, and generate the corresponding first size table. For example, the structural information of the size table to be generated may include the number of tables in the first size table to be generated, the dimensions of each table, the table style, etc.

[0067] It should be noted that when the client responds to the size viewing event of the target product, it can send the product identifier of the target product to the server. If the server has stored the first size table corresponding to the original size information of each product, it can directly obtain the first size table of the target product based on the product identifier and send it back to the client for display. If the server does not store the first size table corresponding to the original size information of the target product, the server can obtain the original size information of the target product based on the product identifier and convert the original size information into a structured first size table.

[0068] One implementation involves pre-obtaining the original size information of each product on the content application platform, generating a corresponding first size table, and storing it. In other words, the content application platform pre-converts the original size information of each product to obtain a structured first size table, and stores the first size table corresponding to the original size information of each product.

[0069] Another approach is to respond to the size viewing event of the target product, obtain the original size information of the target product, and generate the corresponding first size table; that is, when the user views the size of the target product, the content application platform converts the original size information of the target product to obtain a structured first size table and displays it to the user.

[0070] Example, Figure 3 This is a schematic diagram illustrating the generation process of a first size chart for a target product according to one embodiment of this specification, as shown below. Figure 3 As shown, assuming the target product is product 1, the original size information of product 1, "size", will be used. Figure 1 The target extraction model takes the second target prompt word as input and outputs "size". Figure 1 The structured size data is used to construct a first target prompt word based on the structured size data and a first preset prompt word template. The first target prompt word is then input into the target generation model to obtain a first size table corresponding to product 1. This first size table is a two-dimensional product size table. Assuming the target product is product 2, the original size information of product 2 is used... Figure 2 The target extraction model takes the second target prompt word as input and outputs "size". Figure 2 The structured size data of "" is used to construct a first target prompt word based on the structured size data and the first preset prompt word template. The first target prompt word is then input into the target generation model to obtain the first size table corresponding to product 2. This first size table is a two-dimensional product size table. For example... Figure 2 As shown, although "size" Figure 1 "and size" Figure 2 "Although the styles of the size charts are different, after processing by the target extraction model and the target generation model, the first size chart displayed is a two-dimensional product size chart with a fixed structure and style, which realizes the structured processing of the size chart."

[0071] In this embodiment of the specification, in response to a size viewing event of a target product, in addition to displaying a first size table of the target product, a table conversion control and a size recommendation information block can also be displayed. The table conversion control is used to switch the language and / or size unit of the first size table, and the size recommendation information block is used to display recommended sizes for the current account, which are determined based on the account information of the current account and the first size table.

[0072] Specifically, the account information for the current account can include at least one of the following: body type characteristics, historical purchase behavior, and account review information. Body type characteristics refer to the current account's body-related parameters, such as height / weight, bust / waist / hip measurements, and shoulder width, which can provide a more accurate basis for size recommendations. Historical purchase behavior refers to the current account's size selection records when purchasing goods in the past, such as the clothing and shoe sizes that the account frequently buys. By analyzing historical purchase behavior, we can understand the account's size preferences, thus providing a more reliable basis for size recommendations. Account review information refers to the reviews posted by the current account after purchasing goods in the past. For example, if the current account purchased size M clothing and reviewed it as "too loose," the account review information can provide a more reliable basis for size recommendations by understanding the current account's genuine feedback on the relevant sizes.

[0073] Example, Figure 4 This is a schematic diagram of a product details page for a target product provided in one embodiment of this specification, such as... Figure 4 As shown, the product details page can display the first size chart of the target product, a table conversion control, and a size recommendation information block. The size recommendation information block displays the recommended size for the current account, "We recommend you choose size M".

[0074] In one optional implementation of this embodiment, in response to a size viewing event of the target product, a first size chart of the target product is displayed, including:

[0075] In response to the size viewing event of the target product, the first size table in the target language is obtained and displayed. The first size table is generated based on the original size information of the target product, through structural prompts and target generation model to generate a candidate size table, and the candidate size table is obtained by switching to the target language, which is the language of the current account's region.

[0076] It should be noted that the language of the candidate size table generated based on the original size information of the target product and through structural prompts and the target generation model is consistent with the language of the original size information. That is, if the original size information uploaded by the merchant is in Chinese, the candidate size table generated through structural prompts and the target generation model will also be in Chinese. However, different regions have different language habits, so the language in the candidate size table generated by the target generation model can be translated into the language of the current account's region before displaying the first size table in that target language to the user.

[0077] In practice, in response to a size viewing event for a target product, the client can send the product identifier and the account identifier of the current account to the server. The server can then obtain the account's geographic information based on the account identifier. This geographic information indicates the region where the account is located, such as the account's IP address (Internet Protocol Address), shipping address, and entered region. The server determines the current account's region based on this geographic information and the corresponding target language. Using a translation tool, it translates the language in the candidate size list generated based on the target size generation model into the corresponding target language, obtaining the first size list, which is then sent back to the client for display. Alternatively, the server can pre-store first size lists in different language versions and, in response to the client's size viewing event for a target product, return the first size list corresponding to the language of the current account's region.

[0078] Example, Figure 5 This is a schematic diagram of a size chart in different languages ​​provided in one embodiment of this specification, as shown below. Figure 5 As shown, if the current account's IP address or shipping address indicates a user using a Chinese region, the first size chart in Chinese will be displayed; if the IP address or shipping address indicates a user using an English region, the first size chart in English will be displayed; if the IP address or shipping address indicates a user using a Traditional Chinese region, the first size chart in Traditional Chinese will be displayed.

[0079] In this embodiment of the specification, taking into account the language differences of users in different regions, the language version of the first size chart is intelligently identified and switched. Based on the current account's location, the first size chart corresponding to the language of that region is directly displayed. The displayed first size chart is closer to the user's language habits, improves the user experience, and meets the needs of users in different regions.

[0080] In an optional implementation of this embodiment, the method further includes:

[0081] In response to the switching operation of the media content display area on the product details page of the target product, the target media content of the target product is displayed in the media content display area. The target media content is determined based on the size information of the current account.

[0082] Specifically, the product details page of the target product includes a media content display area, which can display media content related to the target product, such as model try-on photos, try-on videos, and size charts.

[0083] It should be noted that the media content display area can provide a switching function. When a user triggers the switching operation of the media content display area, the target media content determined based on the size information of the current account can be switched to be displayed in the media content display area. In other words, the target media content with a similar size to the current account can be provided for the user's reference.

[0084] Example, Figure 6 This is a schematic diagram illustrating the switching of the media content display area in a product details page, provided by one embodiment of this specification. Figure 6 As shown, the media content area on the product details page initially displays image 1 of a model wearing the product, who is 170cm tall, weighs 50kg, has a bust of 80cm, a waist of 60cm, and hips of 85cm. Assuming a user triggers a switching action in this media content display area, such as clicking, swiping left, or swiping right, and determines the current account's size information (e.g., height 165cm, weight 80kg, bust of 98cm, waist of 78cm, hips of 103cm), the media content area then switches to displaying image 2 of a model wearing the product, who is 167cm tall, weighs 82kg, has a bust of 97cm, a waist of 75cm, and hips of 100cm.

[0085] Another example, Figure 7 This is a schematic diagram illustrating the switching of the media content display area in another product details page, provided in one embodiment of this specification. Figure 7 As shown, the media content area on the product details page of the target product initially displays the first size chart. Assuming the user triggers a switching operation in this media content display area, such as clicking, swiping left or right, and determines the current account's size information as height 165cm / weight 70kg, chest 90, waist 75, hip 98, the media content area will then switch to display a size chart that is suitable for the current account, such as marking the recommended size "XL" for the current account in the first size chart.

[0086] In this embodiment of the specification, the media content area can switch to display target multimedia content that matches the user's size, provide media content references similar to the current user, provide media content that better meets the user's expectations, and be closer to the user's actual body shape, helping the user to more accurately understand the effect of the target product on themselves, thereby improving the accuracy and efficiency of size selection and enhancing user stickiness.

[0087] In one optional implementation of this embodiment, the target generation model is trained in the following manner:

[0088] Obtain a first training sample set, wherein the first training sample set includes sample size information of at least two styles, the sample size information corresponds to a size chart label, and the sample size information of at least two styles includes sample size images of different styles and / or sample size description text of different formats.

[0089] The first sample prompt word is constructed based on the first sample size information and the first preset prompt word template. The first sample prompt word is input into the model to be trained to obtain the output predicted size table. The first sample size information is any one of the first training sample sets.

[0090] Based on the size table labels of the predicted size table and the size information of the first sample, the model parameters of the generator model to be trained are adjusted, and the process returns to continue to build the first sample prompt word based on the size information of the next sample and the first preset prompt word template. The generator model to be trained is trained until the training stopping condition is met, and the target generator model is obtained after training is completed.

[0091] It should be noted that a large number of sample size images of different styles and / or sample size description texts of different formats can be collected to construct a first training sample set. This first training sample set includes sample size information of at least two styles, and each sample size information is labeled with a corresponding size table label, which serves as the true label result for the sample size information. This sample size information can be sample size images and / or sample size descriptions.

[0092] In practice, the first sample prompt words can be constructed based on the first sample size information and the first preset prompt word template. The first sample prompt words are then input into the model to be trained to obtain the output predicted size table.

[0093] Specifically, arbitrary size information can be selected from the first training sample set. Based on this size information and a pre-defined prompt word template, a first sample prompt word is constructed. This prompt word is then input into the generator model to be trained, yielding a corresponding predicted size table. This predicted size table represents the prediction result generated by the generator model processing the size information according to the prompt word. Based on the predicted size table and the size table labels corresponding to the first sample size information, the model's prediction loss can be determined. Based on this loss, the model parameters are adjusted. The generator model continues to be trained by constructing a new prompt word based on the next size information from the first training sample set and the pre-defined prompt word template, inputting it into the generator model, and continuing training until the training stopping condition is met, resulting in a completed target generator model.

[0094] In one possible implementation, the training stopping condition can be determined solely based on the relationship between the model's predicted loss and the loss threshold. Specifically, if the model's predicted loss is greater than or equal to the loss threshold, it indicates a significant difference between the predicted size chart and the size chart labels (the model's prediction result differs greatly from the true label result), indicating poor extraction and generation capabilities of the model to be trained. In this case, it can be determined that the training stopping condition has not been met. The model parameters of the model to be trained are adjusted, and training continues until the model's predicted loss is less than the loss threshold. This indicates that the significant difference between the predicted size chart and the size chart labels is smaller (the model's prediction result is closer to the true label result), thus confirming that the training stopping condition has been met, and the trained target generation model is obtained. The loss threshold is a critical value for the loss, and its specific value is selected based on actual circumstances; this embodiment does not impose any limitations on this.

[0095] In another possible implementation, besides comparing the relationship between the model's predicted loss and the loss threshold, the number of iterations can also be used to determine whether the training stopping condition has been met. Specifically, if the model's predicted loss is greater than the loss threshold, it can be further determined whether the current iteration count has reached the preset number of iterations. If the current iteration count has not reached the preset number of iterations, it can be determined that the training stopping condition has not been met. The model parameters of the generator model to be trained can be adjusted, and training can continue until the preset number of iterations is reached. At this point, the training stopping condition is determined, iteration is stopped, and the trained target generator model is obtained. The preset number of iterations is set according to the actual situation. When the preset number of iterations is reached, it means that the generator model has been trained sufficiently, the model's prediction result is very close to the true label result, and training can be stopped.

[0096] In the embodiments of this specification, based on the predicted size table output by the model to be trained for the size information of the first sample, and the size table labels corresponding to the size information of the first sample, the model prediction loss can be determined, and the difference between the model prediction result and the true label result of the model to be trained can be intuitively shown. Based on this difference, the model to be trained can be trained in a targeted manner, which can effectively improve the training rate and training effect of the model to be trained. The target generation model obtained by training can receive the original size information of any sample and generate a size table with the corresponding structure according to the requirements of the prompt words of the first sample. The target generation model completed by training can realize efficient parsing of information and accurate structured output, providing users with a clear and uniform size display.

[0097] In one optional implementation of this embodiment, the first sample prompt word is constructed based on the first sample size information and the first preset prompt word template, including:

[0098] Input the second target prompt and the size information of the first sample into the target extraction model to obtain the structured size data of the size information of the first sample;

[0099] Based on the structured size data of the first sample size information and the first preset prompt word template, the first sample prompt word is constructed.

[0100] The target extraction model is a multimodal large model that can identify unstructured raw size information, such as size images and size description text, extract structured size data that conforms to table features, and return formatted structured size data, such as JSON format.

[0101] In practice, the second target prompt word and the first sample size information can be input into the target extraction model to obtain structured size data of the first sample size information. That is, the trained target extraction model extracts the structured size data from the first sample size information, and then fills this structured size data into the first preset prompt word template to construct the corresponding first sample prompt word. This prompt word is then input into the generation model to be trained. In other words, the trained target extraction model can be used to extract structured size data from sample size information for training the generation model.

[0102] In another implementation, the second target prompt word and the size information of the first sample can be input into the extraction model to be trained to obtain the predicted structured size data of the first sample size information. This predicted structured size data is then filled into the first preset prompt word template to construct the corresponding first sample prompt word. The first sample prompt word is then input into the generation model to be trained to obtain the predicted size table. Based on the predicted size table and the size table labels of the first sample size information, the model parameters of the extraction model and the generation model to be trained are adjusted, and the extraction model and the generation model to be trained are trained simultaneously. That is, the extraction model and the generation model to be trained are trained simultaneously.

[0103] As an example, the second target prompt constructed based on the structured size data extraction method could be: "#Role: Data Extraction Assistant. ##Profile - Description: Extract structured data that conforms to table features from size images and return it as a JSON array in two-dimensional format. ##Skills: 1. Possess image recognition capabilities, able to accurately identify and parse table content in images; 2. Possess data structuring capabilities, able to convert extracted table data into a two-dimensional array in JSON format; 3. Possess multi-table processing capabilities, able to extract multiple tables from a single image and return them as an array list; 4. Possess data cleaning and formatting capabilities, ensuring that table data is neat and error-free. ##Rules: 1. Identify and extract all table data in the image, storing each table independently as a two-dimensional array; 2. Ensure the table headers are complete and the row and column data are aligned." 1. Ensure all valid information is included; 2. Output format should be JSON, with an example data format as follows…; 3. If the image contains no table data, return an empty array `[]`; 4. Output only table-related data, without other image text or background information. ## Workflow: 1. Parse the input image and check if it contains table data; 2. Identify the table boundaries, headers, column fields, and data rows one by one, and convert them into a two-dimensional array; 3. Process multiple table data to ensure the data format is neat and correct; 4. Return all extracted table data in JSON format. ## Input - An image containing clothing size information, “Example Image”. ## Output - Extracted table data, format example:…. ## Initialization: Please provide an image link containing clothing size information, and I will extract the table data for you and return the result in JSON format.

[0104] Inputting the second target prompt and the first sample size information into the target extraction model yields structured size data of the first sample size information. Filling this structured size data into the first preset prompt template generates the first sample prompt: "#Role: Size Chart Data Analysis Expert.##Task - From the given two-dimensional array 'Structured Size Data,' identify and extract the type of size chart, and output size chart data that distinguishes different types according to the standard structure.##1. Understand the following size chart type descriptions: Basic Size Chart: Contains sizes and specific garment length parameters (such as waist circumference, trouser length, chest circumference, etc.); Size Recommendation Chart:" Includes size and user body parameters (such as height, weight, age, etc.); Model try-on table: includes size and model's experience wearing that size (such as model name, tried-on size, height, weight, etc.). ##2. Processing Requirements: If the input contains information from multiple size table types, they should be extracted separately and output according to the corresponding format; ensure that the output format strictly meets the requirements. The basic size table is a two-dimensional array, the size recommendation table format has a maximum of two columns of parameters plus size, and the model try-on table contains fixed columns. ##3. JSON Output Structure: Output content according to JSON format, which should conform to the following structure, where the table header format is "parameter name (unit)", parameter... Numerical values ​​that are ranges must be converted to a symbolic representation limited to "-", "<", and ">". Each array value must be a complete two-dimensional array and be deserializable into a map. ##4. Formatting Rules: Unit Formatting: Output values ​​should have their units removed, and the header should specify the unit (e.g., cm, kg); Priority Matching: For size recommendation tables, prioritize matching key columns such as "height" and "weight". ##5. Special Requirements: If the header contains synonyms or similar text, priority matching should be based on column names; if the size recommendation table has multiple columns and more than two parameters, only the column names from left to right should be retained. Two columns plus a size column. ## Workflow: 1. Receive input: formatted text content; 2. Determine the size table type based on the text content and extract relevant data; 3. If it doesn't match any size table type, return an empty list; 4. The header and form values ​​may not contain units. Consider the reasonableness of the number range, determine the corresponding length or weight unit, and add the unit to the header of each column according to the format, but do not perform unit conversion; 5. Output compliant JSON data, without containing other text. ## Example 1 - Input: ... ## Example 1 - Output: ... ### The following is the size table content to be processed: ...

[0105] In the embodiments of this specification, structured size data can be extracted from sample size information using a target extraction model, converting various unstructured sample size information into structured size data, and then generating a structured size table. The target extraction model can integrate the processing capabilities of multimodal data such as images and text, accurately identify and parse unstructured size information, generate corresponding structured size data, and achieve efficient parsing and accurate structured output of unstructured size information.

[0106] Of course, in actual implementation, a large model (such as an extraction and generation model) can be used to simultaneously extract structured size data and generate structured size tables. For example, a structure prompt word can be constructed, which can indicate the extraction method of structured size data in the original size information and the structural information of the size table to be generated. The original size information and the structure prompt word are input into the large model, which can process the input original size information according to the instructions of the structure prompt word, extract the structured size data, and process the structured size data according to the instructions of the structure prompt word to generate a size table with the corresponding structure. This embodiment of the specification does not limit this.

[0107] In one optional implementation of this embodiment, after obtaining the trained target generation model, the method further includes:

[0108] Determine the prediction error of the size information of each sample in the first training sample set;

[0109] Based on the prediction error, add the corresponding error correction constraint to the first preset prompt word template to generate an updated prompt word template;

[0110] Obtain a second training sample, construct a second sample prompt word based on the sample size information and update prompt word template in the second training sample set, and fine-tune and train the target generation model to obtain a first update generation model. The second training sample may be the same as or different from the first training sample.

[0111] It should be noted that after constructing the first sample prompt words based on the first training sample set and combining the first preset prompt word template, and training the target generation model to be trained, the first preset prompt word template can be further iterated and optimized to further fine-tune the target generation model.

[0112] In practice, the prediction error for the size information of each sample in the first training sample set can be obtained. Based on this prediction error, corresponding error correction constraints are added to the first preset prompt word template to generate an updated prompt word template. Here, the prediction error refers to the difference between the predicted size table and the size table label corresponding to the size information of each sample, and the error correction constraint refers to the constraint condition that can reduce the difference.

[0113] Specifically, based on the prediction error between the predicted size table and size table label corresponding to the size information of each sample, the reasons for the difference in size information of each sample can be analyzed, the proportion of each reason for the difference can be counted, the reasons for the difference that have a proportion higher than the threshold can be selected to be corrected, the corresponding error correction constraints can be generated, the update prompt word template can be generated, and then the optimization training can be carried out.

[0114] As an example, suppose the first training sample includes size information of 100 samples, and the difference in size information of 70 samples is due to incorrect chest circumference extraction. In this case, an error correction constraint "extract chest circumference value and label unit format" can be added to generate an update prompt word template.

[0115] It should be noted that after generating the updated prompt word template, the target generation model can be fine-tuned and trained in conjunction with the second training sample to obtain the first updated generation model. The specific fine-tuning and training process is similar to the training process described above, and will not be repeated here in the embodiments of this specification.

[0116] In the embodiments of this specification, based on the prediction error of the size information of each sample in the first training sample set, the first preset prompt word template can be iteratively optimized, constraints can be added in a finer granular manner, the model's ability to accurately parse structured size data can be improved, and the accuracy of generating structured size tables can be continuously improved.

[0117] In one optional implementation of this embodiment, after obtaining the trained target generation model, the method further includes:

[0118] Determine the prediction error of the size information of each sample in the first training sample set;

[0119] The negative training samples in the first training sample set are determined based on the prediction error.

[0120] Determine the error factors of the negative training samples, and update the first training sample set based on the error factors to obtain the third training sample set;

[0121] Based on the sample size information in the third training sample set and the first preset prompt word template, a third sample prompt word is constructed, and the target generation model is optimized and trained to obtain a second updated generation model.

[0122] It should be noted that after obtaining the trained target generation model, the training samples can be further iterated and optimized to further fine-tune the target generation model.

[0123] In practice, the prediction error of the size information of each sample in the first training sample set can be obtained. This prediction error can indicate the prediction loss value of the size information of each sample. If the prediction loss value is greater than the threshold, the size information of the sample is determined as a negative sample. The error factor of the negative training sample is determined. This error factor is used to indicate the reason for the occurrence of the negative sample. For example, the error factor can be that the number of training samples is small, the quality is poor, or the richness is low. Then, based on the error factor, the first training sample set can be updated to obtain the third training sample set. Then, the target generation model is tuned and trained based on the updated third training sample set.

[0124] As an example, suppose the predicted loss value of a sample size information is greater than the threshold, the sample size information is a negative sample, and its error factor is that there are few samples in this scenario. At this time, a large number of samples in this scenario can be obtained and added to the original first training sample set to obtain a third training sample set with rich scene samples.

[0125] In the embodiments of this specification, based on the prediction error of the size information of each sample in the first training sample set, the error factors can be analyzed, the training sample set can be iteratively optimized, and the model's ability to accurately parse structured size data can be improved, thereby continuously improving the accuracy of generating structured size tables.

[0126] Step 104: In response to the trigger operation of the table conversion control, switch the first size table to the second size table; wherein the second size table is the table after switching the language and / or size units of the first size table.

[0127] It should be noted that the system can display a first size chart for the target product, as well as a table conversion control. This table conversion control is used to switch the language and / or size unit of the first size chart. If the currently displayed first size chart does not conform to the user's language and / or size unit usage habits, the user can trigger the table conversion control to switch the language and / or size unit of the first size chart and display the second size chart after the switch.

[0128] Specifically, the second size table can be output simultaneously with the first size table based on the same target generation model. That is, the first target prompt word is input into the target generation model to obtain the first size table and the second size table in different languages ​​and / or different size units; or, the first size table can be generated first, and then the language and / or size unit conversion can be performed on the first size table to obtain the corresponding second size table.

[0129] In actual implementation, the table conversion control includes a language conversion control and / or a size unit conversion control. When the client detects a trigger operation of the language conversion control, it can send the trigger operation to the server. The server determines the target language to be switched to corresponding to the trigger operation, queries the database for the second size table of the target language, and sends this second size table back to the client. The client then displays the second size table in the target language. Similarly, when the client detects a trigger operation of the size unit conversion control, it can send the trigger operation to the server. The server determines the target size unit to be switched to corresponding to the trigger operation, queries the database for the second size table of the target size unit, and sends this second size table back to the client. The client then displays the second size table in the target size unit.

[0130] Example, Figure 8 This is a schematic diagram illustrating the switching of a size chart according to one embodiment of this specification, such as... Figure 8 As shown, the product details page of the target product displays a first size chart, a table conversion control, and a size recommendation information block. The table conversion control includes a language conversion control - "Simplified Chinese", "Traditional Chinese", and "English", and a size unit switching control - "kg" and "pound".

[0131] Assume the first size chart is currently displayed in "Simplified Chinese" and the unit of measurement is "kg". The size recommendation information block states "We recommend you wear size L". If the user clicks "English", the language conversion control is triggered, and the server converts "Simplified Chinese" to "English" in the first size chart. If the user clicks "pounds", the unit of measurement conversion control is triggered, and the server converts "kg" to "pounds" in the first size chart.

[0132] In the embodiments described in this specification, size unit switching, multi-language adaptation, and intelligent recommendation are supported. The modular interactive framework integrates structured size charts, user profile management (i.e., body shape characteristics), and size recommendation functions, presenting them with visualized data to enhance user trust and decision-making efficiency, making it easier for users to quickly select suitable sizes and assisting in user decision-making.

[0133] In an optional implementation of this embodiment, the method further includes:

[0134] In response to the triggering operation of the size recommendation information block, the size recommendation page is displayed, which displays at least one of the following: the recommended size for the current account, the size information of the reference account, size media content, and product pattern information.

[0135] It should be noted that when a user triggers the size recommendation information block, a size recommendation page will be displayed. This size recommendation page is a size assistant that can help users select sizes. The size recommendation page displays more detailed size recommendation information, such as at least one of the following: recommended sizes for the current account, size information for reference accounts, size media content, and product pattern information.

[0136] The recommended size for the current account refers to a size determined based on at least one of the following account information: body type, historical purchase behavior, and account reviews. Reference account size information indicates the size information purchased by other users similar to the current account's size information. Size media content refers to try-on videos or images similar to the current account's size information, or try-on videos or images with multiple different size information. Product fit information refers to the fit of the target product, such as loose, slim, or tight.

[0137] In the embodiments described in this manual, when a user triggers the size recommendation information block, a size recommendation page can be displayed, providing richer size recommendation content and offering users more comprehensive size references. This facilitates users in verifying the recommendation logic (such as viewing the algorithm basis, similar references, etc.), reducing decision-making anxiety caused by information asymmetry, assisting users in making purchasing decisions, and greatly improving the accuracy and efficiency of size selection.

[0138] In one optional implementation of this embodiment, the account information of the current account includes at least one of body shape characteristics, historical purchase behavior, and account rating information; displaying the recommended size for the current account on the size recommendation page includes at least one of the following:

[0139] The size recommendation page displays the first recommended size for the current account. The first recommended size is obtained by matching the body shape characteristics filled in by the current account in the first size table.

[0140] The size recommendation page displays the second recommended size for the current account. The second recommended size is obtained by matching the estimated product size corresponding to the current account with the first size table. The estimated product size is determined based on the current account's historical purchase behavior.

[0141] The size recommendation page displays the third recommended size for the current account. The third recommended size is obtained by matching the estimated body shape characteristics of the current account with the first size table. The estimated body shape characteristics are determined based on the current account's historical purchase behavior and / or account rating information.

[0142] It should be noted that the account information for the current account can include at least one of the following: body type characteristics, historical purchase behavior, and account review information. Body type characteristics refer to the current account's body-related parameters, such as height / weight, bust / waist / hip measurements, and shoulder width, which can provide a more accurate basis for size recommendations. Historical purchase behavior refers to the current account's size selection records when purchasing goods in the past, such as the clothing and shoe sizes that the account frequently buys. By analyzing historical purchase behavior, we can understand the account's size preferences, thus providing a more reliable basis for size recommendations. Account review information refers to the reviews posted by the current account after purchasing goods in the past. For example, if the current account purchased size M clothing and reviewed it as "too loose," the account review information can provide a true feedback on the current account regarding the relevant size, thus providing a more reliable basis for size recommendations.

[0143] In practice, if the account information includes the body shape characteristics entered by the current account holder, the system matches these characteristics against the first size table to determine the corresponding first recommended size. If the account information includes historical purchase behavior, the system determines the estimated size of the target product based on the historical purchase behavior, and matches this estimated size against the first size table to determine the corresponding second recommended size. If the account information includes historical purchase behavior and / or user review information, the system determines the estimated body shape characteristics of the current account based on the historical purchase behavior and / or user review information, and matches this estimated body shape characteristics against the first size table to determine the corresponding third recommended size.

[0144] It should be noted that if the recommended size for the current account is determined based on different information, then each recommended size and the corresponding reason for the recommendation can be displayed separately.

[0145] Continuing with the previous example, such as Figure 4 As shown, assuming a user triggers this size recommendation information block, a size recommendation page will be displayed. This page shows the product fit as loose; the recommended size for the current account as: based on your body type, we suggest you wear size L; based on your past purchases and / or account reviews, size L may be too loose, we suggest you wear size M; reference account size information: reference users 1-n, each with their height, weight, and purchased sizes; size media content: model try-on images of pear-shaped, slightly overweight, and slender figures.

[0146] In the embodiments of this specification, at least one of the following can be combined: the body shape characteristics of the current account, historical purchase behavior, and account evaluation information. These can be matched in a structured first size table to generate a highly accurate recommended size, reduce the risk of decision-making bias, and achieve an upgrade from "static matching of single data" to "dynamic matching of multiple data fusion". This significantly improves the efficiency and accuracy of users' size decisions, enhances the efficiency of users' shopping decisions, and helps users find the most suitable size for themselves faster and more accurately.

[0147] In one optional implementation of this embodiment, the size recommendation page includes a body shape feature update control; the method further includes:

[0148] In response to the triggering operation of the body shape feature update control, display the body shape feature update page;

[0149] The system retrieves the updated body shape features entered on the body shape feature update page and displays updated size recommendations for the current account based on these features. The updated size recommendations are obtained by matching the updated body shape features against the first size table.

[0150] It should be noted that the size recommendation page can also include a body shape feature update control. When the user triggers this control, the body shape feature update page will be displayed. The user can re-enter updated body shape features on this page, and the client can send the updated body shape features to the server. The server then uses these updated features to match and obtain updated size recommendations from the first size table, and sends the updated recommendations back to the client. The client then displays the updated size recommendations on the size recommendation page. In this way, users can dynamically update their body shape features, achieving real-time dynamic size recommendations.

[0151] Continuing with the previous example, such as Figure 4 As shown, the size recommendation page includes a body shape feature update control. If a user clicks this control, the body shape feature update page is displayed. The user can re-enter their updated body shape features on this page and click "Confirm." This updated body shape feature is then sent to the server. The server returns an updated size recommendation based on this updated body shape feature, matching it against the first size table. This update then updates the displayed size recommendation page. Figure 4 As shown, the size recommendation page currently displays the product fit as loose; the recommended size for the current account is: based on your current body type, we recommend size M; based on your past purchases and / or account reviews, we also recommend size M, which is relatively loose on the upper body; reference account size information: reference users 1-n, each reference user's height, weight, and purchased sizes; size media content: model try-on photos of pear-shaped, slightly overweight, and slender figures.

[0152] This specification provides a product size recommendation method. Based on the requirements of structural prompts, a target generation model generates a first size table corresponding to the original size information of the target product. This unifies the diverse, poorly formatted, and redundant original size information into a first size table with a corresponding structure, providing users with a clear and accurate size table. Based on the current account information and the first size table, the recommended size for the current account is determined and displayed in the size recommendation information block, thus providing accurate size recommendations without cumbersome operations. Furthermore, a table conversion control can be displayed, providing the ability to switch the language and / or size units of the first size table, offering a size table that better suits the user's actual language and / or size habits. This allows users to quickly find the most suitable size, assisting in their purchase decisions and greatly improving the accuracy and efficiency of size selection.

[0153] See Figure 9 , Figure 9 A flowchart of a method for generating a product size chart according to an embodiment of this specification is shown, specifically including the following steps.

[0154] Step 902: Obtain the original size information of the target product, wherein the original size information is the original size image and / or the original size description text.

[0155] Step 904: Input the original size information of the target product and the second target prompt word into the target extraction model to obtain the structured size data of the original size information, and construct the first target prompt word based on the structured size data of the original size information and the first preset prompt word template.

[0156] Step 906: Input the first target prompt word into the target generation model to obtain the first size table of the target product. The target generation model is trained using sample size information of at least two styles.

[0157] It should be noted that the above is an illustrative scheme of a product size chart generation method in this embodiment. The technical solution of this product size chart generation method belongs to the same concept as the technical solution of the product size recommendation method described above. For details not described in detail in the technical solution of the product size chart generation method, please refer to the description of the technical solution of the product size recommendation method described above.

[0158] This specification provides a method for generating a product size chart. Following the prompts, a target generation model generates a first size chart corresponding to the original size information of the target product. Utilizing the data processing capabilities of the target generation model, the original size information is unified into a first size chart with a corresponding structure, providing users with a clear and accurate size chart to assist in their purchasing decisions and significantly improve the accuracy and efficiency of size selection.

[0159] See Figure 10 , Figure 10 A flow chart of a product size recommendation method according to an embodiment of this specification is shown, as follows: Figure 10 As shown, the client is mainly responsible for user interaction and data collection, while the server is located on a remote server or cloud service provider and is responsible for data processing, storage, and interaction with external systems (such as databases, third-party services, etc.).

[0160] On the server side, a first training sample set is obtained, which includes sample size information in at least two styles, each with a corresponding size chart label. The sample size information in the at least two styles includes sample size images with different styles and / or sample size description text with different formats. The second target prompt word and the first sample size information are input into the target extraction model to obtain structured size data of the first sample size information. Based on the structured size data of the first sample size information and a first preset prompt word template, a first sample prompt word is constructed. The first sample prompt word is input into the model to be trained to obtain the output predicted size chart, where the first sample size information is any one of the first training sample sets. Based on the predicted size chart and the size chart labels of the first sample size information, the model parameters of the model to be trained are adjusted, and the process returns to continue constructing the first sample prompt word based on the next sample size information and the first preset prompt word template. The model to be trained is trained until the training stopping condition is met, and the trained target generation model is obtained.

[0161] On the server side, the original size information of the target product is obtained, wherein the original size information is the original size image and / or the original size description text; the original size information and the second target prompt word are input into the target extraction model to obtain the structured size data of the original size information; based on the structured size data and the first preset prompt word template, the first target prompt word is constructed; the first target prompt word is input into the target generation model to generate the corresponding first size table.

[0162] When the client detects a size viewing event for the target product, it sends a size viewing request for the target product to the server.

[0163] On the server side, in response to the size view of the target product, the server retrieves the first size chart of the target product; and retrieves the account information of the current account. Based on the account information and the first size chart, the server determines the recommended size for the current account; and returns the first size of the target product and the recommended size for the current account to the client.

[0164] The client displays the first size chart of the target product, a table conversion control, and a size recommendation information block, in which the recommended sizes for the current account are displayed.

[0165] The client, in response to the triggering operation of the table conversion control, sends a switching request to the server.

[0166] On the server side, in response to the switching request, the corresponding second size table is queried and the second size table is sent back to the client. The second size table is obtained by switching the language and / or size unit of the first size table.

[0167] The client will switch the displayed first size chart to the second size chart.

[0168] The client, in response to the triggering operation of the size recommendation information block, sends a request to the server to retrieve the size recommendation page.

[0169] On the server side, in response to the request to retrieve the size recommendation page, the recommended size for the current account, the size information of the reference account, the size media content, and the product pattern information are obtained and then fed back to the client.

[0170] The client displays recommended sizes for the current account, reference account size information, size media content, and product fit information on the size recommendation page.

[0171] This specification provides a product size recommendation method. It generates a first size table corresponding to the original size information of the target product through a target generation model. This unifies the complex, diverse, disorganized, and redundant original size information into a first size table with a corresponding structure, providing users with a clear and accurate size table. Based on the current account's information and the first size table, it determines the recommended size for the current account and displays it in the size recommendation information block, thus providing accurate size recommendations without cumbersome operations. Furthermore, it can display a table conversion control, providing the ability to switch the language and / or size units of the first size table, offering a size table that better suits the user's actual language and / or size habits. This allows users to quickly find the most suitable size, assisting in their purchasing decisions and greatly improving the accuracy and efficiency of size selection.

[0172] Corresponding to the above method embodiments, this specification also provides embodiments of a product size recommendation device. Figure 11 A schematic diagram of a product size recommendation device according to one embodiment of this specification is shown. Figure 11 As shown, the device includes:

[0173] The first display module 1102 is configured to respond to the size viewing event of the target product and display the first size table, table conversion control and size recommendation information block of the target product. The first size table is obtained by inputting the first target prompt word into the target generation model. The first target prompt word is constructed based on the original size information and the first preset prompt word template. The size recommendation information block displays the recommended size for the current account. The recommended size is determined based on the account information of the current account and the first size table.

[0174] The switching module 1104 is configured to switch the first size table to the second size table in response to a trigger operation of the table conversion control; wherein the second size table is a table after the language and / or size units of the first size table have been switched.

[0175] Optionally, the device further includes a second display module configured to:

[0176] In response to the switching operation of the media content display area on the product details page of the target product, the target media content of the target product is displayed in the media content display area. The target media content is determined based on the size information of the current account.

[0177] Optionally, the first display module 1102 is further configured as follows:

[0178] In response to the size viewing event of the target product, the first size table in the target language is obtained and displayed. The first size table is generated based on the original size information of the target product, through structural prompts and target generation model to generate a candidate size table, and the candidate size table is obtained by switching to the target language, which is the language of the current account's region.

[0179] Optionally, the device also includes a third display module, configured as follows:

[0180] In response to the triggering operation of the size recommendation information block, the size recommendation page is displayed, which displays at least one of the following: the recommended size for the current account, the size information of the reference account, size media content, and product pattern information.

[0181] Optionally, the account information of the current account includes at least one of body shape characteristics, historical purchase behavior, and account rating information; the third display module is further configured to include at least one of the following:

[0182] The size recommendation page displays the first recommended size for the current account. The first recommended size is obtained by matching the body shape characteristics filled in by the current account in the first size table.

[0183] The size recommendation page displays the second recommended size for the current account. The second recommended size is obtained by matching the estimated product size corresponding to the current account with the first size table. The estimated product size is determined based on the current account's historical purchase behavior.

[0184] The size recommendation page displays the third recommended size for the current account. The third recommended size is obtained by matching the estimated body shape characteristics of the current account with the first size table. The estimated body shape characteristics are determined based on the current account's historical purchase behavior and / or account rating information.

[0185] Optionally, the size recommendation page includes a body shape feature update control; the device also includes an update module configured to:

[0186] In response to the triggering operation of the body shape feature update control, display the body shape feature update page;

[0187] The system retrieves the updated body shape features entered on the body shape feature update page and displays updated size recommendations for the current account based on these features. The updated size recommendations are obtained by matching the updated body shape features against the first size table.

[0188] Optionally, the first display module 1102 is further configured as follows: the first target prompt word is constructed based on the structured size data of the original size information and the first preset prompt word template. The structured size data of the original size information is obtained by inputting the original size information and the second target prompt word into the target extraction model. The second target prompt word is constructed based on the extraction method of the structured size data in the original size information.

[0189] Optionally, the device also includes a training module configured to:

[0190] Obtain a first training sample set, wherein the first training sample set includes sample size information of at least two styles, the sample size information corresponds to a size chart label, and the sample size information of at least two styles includes sample size images of different styles and / or sample size description text of different formats.

[0191] The first sample prompt word is constructed based on the first sample size information and the first preset prompt word template. The first sample prompt word is input into the model to be trained to obtain the output predicted size table. The first sample size information is any one of the first training sample sets.

[0192] Based on the size table labels of the predicted size table and the size information of the first sample, the model parameters of the generator model to be trained are adjusted, and the process returns to continue to build the first sample prompt word based on the size information of the next sample and the first preset prompt word template. The generator model to be trained is trained until the training stopping condition is met, and the target generator model is obtained after training is completed.

[0193] Optionally, the training module is further configured as follows:

[0194] Input the second target prompt and the size information of the first sample into the target extraction model to obtain the structured size data of the size information of the first sample;

[0195] Based on the structured size data of the first sample size information and the first preset prompt word template, the first sample prompt word is constructed.

[0196] Optionally, the training module is further configured as follows:

[0197] Determine the prediction error of the size information of each sample in the first training sample set;

[0198] Based on the prediction error, an error correction constraint is added to the first preset prompt word template to generate an updated prompt word template;

[0199] Obtain a second training sample, construct a second sample prompt word based on the sample size information and update prompt word template in the second training sample set, and fine-tune and train the target generation model to obtain a first update generation model. The second training sample may be the same as or different from the first training sample.

[0200] Optionally, the training module is further configured as follows:

[0201] Determine the prediction error of the size information of each sample in the first training sample set;

[0202] The negative training samples in the first training sample set are determined based on the prediction error.

[0203] Determine the error factors of the negative training samples, and update the first training sample set based on the error factors to obtain the third training sample set;

[0204] Based on the sample size information in the third training sample set and the first preset prompt word template, a third sample prompt word is constructed. The target generation model is then fine-tuned and trained to obtain the second updated generation model.

[0205] This specification provides a product size recommendation device. It generates a first size table corresponding to the original size information of the target product through a target generation model. This unifies the original size information, which is often presented in complex and varied formats, with chaotic layouts and redundant information, into a first size table with a corresponding structure. This provides users with a clear and accurate size table. Based on the current account's information and the first size table, it determines the recommended size for the current account and displays it in the size recommendation information block, thus providing accurate size recommendations without cumbersome operations. Furthermore, it can display a table conversion control, providing the ability to switch the language and / or size units of the first size table, offering a size table that better suits the user's actual language and / or size habits. This allows users to quickly find the most suitable size, assisting in their purchasing decisions and greatly improving the accuracy and efficiency of size selection.

[0206] The above is a schematic scheme of a product size recommendation device according to this embodiment. It should be noted that the technical solution of this product size recommendation device and the technical solution of the product size recommendation method described above belong to the same concept. For details not described in detail in the technical solution of the product size recommendation device, please refer to the description of the technical solution of the product size recommendation method described above.

[0207] Corresponding to the above method embodiments, this specification also provides embodiments of a device for generating product size charts. Figure 12 A schematic diagram of a device for generating a product size chart according to one embodiment of this specification is shown. Figure 12 As shown, the device includes:

[0208] The acquisition module 1202 is configured to acquire the original size information of the target product, wherein the original size information is an original size image and / or original size description text.

[0209] The construction module 1204 is configured to input the original size information of the target product and the second target prompt word into the target extraction model to obtain the structured size data of the original size information, and to construct the first target prompt word based on the structured size data of the original size information and the first preset prompt word template.

[0210] The generation module 1206 is configured to input the first target prompt word into the target generation model to obtain the first size table of the target product, wherein the target generation model is trained by at least two styles of sample size information.

[0211] This specification provides a product size chart generation device. According to the prompt words, a target generation model generates a first size chart corresponding to the original size information of the target product. By utilizing the data processing capabilities of the target generation model, the original size information is unified into a first size chart with a corresponding structure, providing users with a clear and accurate size chart, assisting users in their purchasing decisions, and greatly improving the accuracy and efficiency of size selection.

[0212] Figure 13 A structural block diagram of a computing device according to one embodiment of this specification is shown. The components of the computing device 1300 include, but are not limited to, a memory 1310 and a processor 1320. The processor 1320 is connected to the memory 1310 via a bus 1330, and a database 1350 is used to store data.

[0213] The computing device 1300 also includes an access device 1340, which enables the computing device 1300 to communicate via one or more networks 1360. Examples of such networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 1340 may include one or more of any type of wired or wireless network interface (e.g., a network interface controller (NIC)), such as an IEEE 802.13 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0214] In one embodiment of this specification, the above-described components of the computing device 1300 and Figure 13 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 13 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0215] The computing device 1300 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 1300 can also be a mobile or stationary server.

[0216] The processor 1320 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described product size recommendation method or product size table generation method.

[0217] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device belongs to the same concept as the above-described product size recommendation method or product size table generation method. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the above-described product size recommendation method or product size table generation method.

[0218] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described product size recommendation method or product size table generation method.

[0219] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the above-described method for recommending product sizes or generating product size tables. Details not described in detail in the technical solution of the storage medium can be found in the description of the technical solution of the above-described method for recommending product sizes or generating product size tables.

[0220] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described product size recommendation method or product size table generation method.

[0221] The above is an illustrative example of a computer program according to this embodiment. It should be noted that the technical solution of this computer program belongs to the same concept as the aforementioned product size recommendation method or product size table generation method. Details not described in detail in the computer program's technical solution can be found in the descriptions of the aforementioned product size recommendation method or product size table generation method.

[0222] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0223] Computer instructions include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in computer-readable media can be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0224] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0225] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0226] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for recommending product sizes, characterized in that, include: In response to a size viewing event of a target product, a first size table, a table conversion control, and a size recommendation information block of the target product are displayed. The first size table is obtained by inputting a first target prompt word into a target generation model. The first target prompt word is constructed based on the original size information and a first preset prompt word template. The size recommendation information block displays the recommended size for the current account. The recommended size is determined based on the account information of the current account and the first size table. In response to a trigger operation on the table conversion control, the first size table is switched to a second size table; wherein the second size table is a table after the language and / or size units of the first size table have been switched.

2. The method according to claim 1, characterized in that, The method further includes: In response to a switching operation of the media content display area in the product details page of the target product, the target media content of the target product is displayed in the media content display area, and the target media content is determined based on the size information of the current account.

3. The method according to claim 1, characterized in that, The method of displaying a first size chart of the target product in response to a size viewing event includes: In response to the size viewing event of the target product, a first size table in the target language is obtained and displayed. The first size table is generated based on the original size information of the target product, through structural prompts and a target generation model to generate a candidate size table, and the candidate size table is obtained by switching the candidate size table to the target language, which is the language of the region where the current account is located.

4. The method according to claim 1, characterized in that, The method further includes: In response to the triggering operation of the size recommendation information block, a size recommendation page is displayed, which displays at least one of the following: the recommended size of the current account, the size information of the reference account, the size media content, and the product pattern information.

5. The method according to claim 4, characterized in that, The account information of the current account includes at least one of body shape characteristics, historical purchase behavior, and account rating information; the recommended sizes for the current account are displayed on the size recommendation page, including at least one of the following: The size recommendation page displays the first recommended size for the current account, wherein the first recommended size is obtained by matching the body shape characteristics filled in by the current account in the first size table; The size recommendation page displays the second recommended size for the current account, wherein the second recommended size is obtained by matching the estimated product size corresponding to the current account with the first size table, and the estimated product size is determined based on the current account's historical purchasing behavior; The size recommendation page displays the third recommended size for the current account, wherein the third recommended size is obtained by matching the estimated body shape characteristics corresponding to the current account in the first size table, and the estimated body shape characteristics are determined based on the current account's historical purchase behavior and / or account rating information.

6. The method according to claim 4, characterized in that, The size recommendation page includes a body shape feature update control; the method further includes: In response to the triggering operation of the body shape feature update control, the body shape feature update page is displayed; The updated body shape features input on the body shape feature update page are obtained, and the updated size recommendation for the current account is displayed based on the updated body shape features. The updated size recommendation is obtained by matching the updated body shape features in the first size table.

7. The method according to claim 1, characterized in that, The first target prompt word is constructed based on the original size information and the first preset prompt word template, including: the first target prompt word is constructed based on the structured size data of the original size information and the first preset prompt word template, wherein the structured size data of the original size information is obtained by inputting the original size information and the second target prompt word into a target extraction model, and the second target prompt word is constructed based on the extraction method of the structured size data in the original size information.

8. The method according to claim 1, characterized in that, The target generation model is trained in the following manner: Obtain a first training sample set, wherein the first training sample set includes sample size information of at least two styles, the sample size information corresponds to a size chart label, and the sample size information of at least two styles includes sample size images of different styles and / or sample size description text of different formats. Based on the first sample size information and the first preset prompt word template, a first sample prompt word is constructed. The first sample prompt word is input into the model to be trained to obtain the output predicted size table. The first sample size information is any one of the first training sample sets. Based on the size table labels of the predicted size table and the size information of the first sample, the model parameters of the generator model to be trained are adjusted, and the process returns to continue to construct the first sample prompt word based on the size information of the next sample and the first preset prompt word template, and train the generator model to be trained until the training stop condition is met, and the target generator model is obtained after training is completed.

9. The method according to claim 8, characterized in that, The step of constructing the first sample prompt word based on the first sample size information and the first preset prompt word template includes: Input the second target prompt word and the size information of the first sample into the target extraction model to obtain the structured size data of the size information of the first sample; Based on the structured size data of the first sample size information and the first preset prompt word template, the first sample prompt word is constructed.

10. The method according to claim 8, characterized in that, After obtaining the trained target generative model, the process also includes: Determine the prediction error of the size information of each sample in the first training sample set; Based on the prediction error, an error correction constraint is added to the first preset prompt word template to generate an updated prompt word template; Obtain a second training sample, construct a second sample prompt word based on the sample size information in the second training sample set and the updated prompt word template, and perform optimization training on the target generation model to obtain a first updated generation model, wherein the second training sample is the same as or different from the first training sample.

11. The method according to claim 8, characterized in that, After obtaining the trained target generative model, the process also includes: Determine the prediction error of the size information of each sample in the first training sample set; Based on the prediction error, negative training samples are determined in the first training sample set; Determine the error factors of the negative training samples, and update the first training sample set based on the error factors to obtain a third training sample set; Based on the sample size information in the third training sample set and the first preset prompt word template, a third sample prompt word is constructed, and the target generation model is optimized and trained to obtain a second updated generation model.

12. A method for generating a product size chart, characterized in that, include: Obtain the original size information of the target product, wherein the original size information is an original size image and / or an original size description text; The original size information of the target product and the second target prompt word are input into the target extraction model to obtain the structured size data of the original size information, and the first target prompt word is constructed based on the structured size data of the original size information and the first preset prompt word template. The first target prompt word is input into the target generation model to obtain the first size table of the target product, wherein the target generation model is trained by at least two styles of sample size information.

13. A product size recommendation device, characterized in that, include: The first display module is configured to respond to a size viewing event of a target product by displaying a first size table, a table conversion control, and a size recommendation information block for the target product. The first size table is obtained by inputting a first target prompt word into a target generation model. The first target prompt word is constructed based on the original size information and a first preset prompt word template. The size recommendation information block displays recommended sizes for the current account. The recommended sizes are determined based on the account information of the current account and the first size table. The switching module is configured to switch the first size table to a second size table in response to a trigger operation of the table conversion control; wherein the second size table is a table after switching the language and / or size units of the first size table.

14. A device for generating a product size chart, characterized in that, include: The acquisition module is configured to acquire the original size information of the target product, wherein the original size information is an original size image and / or original size description text. The construction module is configured to input the original size information of the target product and the second target prompt word into the target extraction model to obtain the structured size data of the original size information, and to construct the first target prompt word based on the structured size data of the original size information and the first preset prompt word template; The generation module is configured to input the first target prompt word into the target generation model to obtain a first size table of the target product, wherein the target generation model is trained using sample size information of at least two styles.

15. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the product size recommendation method according to any one of claims 1-11 or the product size table generation method according to claim 12.

16. A computer-readable storage medium, characterized in that, It stores computer-executable instructions, which, when executed by a processor, implement the steps of the product size recommendation method according to any one of claims 1-11 or the product size table generation method according to claim 12.

17. A computer program product, characterized in that, Includes a computer program / instruction that, when executed by a processor, implements the steps of the product size recommendation method according to any one of claims 1-11 or the product size table generation method according to claim 12.

Images