A SINGLE CATALOG AND CATEGORY METHOD

Abstract

Our invention is a single catalog and category method where data can be entered into a seller screen (10) or received from an integrator (20) via an API, enabling the creation of single product catalogs for easy product finding on online shopping platforms, checking the accuracy of product data and assigning products to the correct categories.

Description

1 TARIFF A SINGLE CATALOG AND CATEGORY METHOD Technical Area The invention relates to a single catalog and category method. More specifically, this is done by entering data into a vendor screen or by receiving data from an integrator via an API 5 individual products that can be purchased with the help of online shopping platforms, making it easy to find products on these platforms. Creating catalogs, checking the accuracy of product data, and ensuring products are correct. It relates to a method that enables the assignment of categories. State of the Art Today, shopping habits have rapidly shifted to online platforms with the advancement of technology. It is shifting. People want to be able to quickly access the products they want and find the best prices without going to physical stores. This allows users to make comparisons. Online shopping saves time. In addition to providing this, it offers a wider range of products, easily meeting users' needs. This makes it possible to meet their needs. On the other hand, budget-friendly options are available through campaigns and discounts. E-commerce sites that offer these services make the purchasing process more economical for users. These 15 Transformation provides flexibility to both consumers and businesses, adapting to the dynamics of modern life. It helps them adapt. E-commerce is a digital trading model where goods and services are bought and sold over the internet. These platforms act as a bridge between buyers and sellers, enabling trade to be conducted quickly and efficiently. It enables this to happen. On these platforms, sellers can showcase their products, including pricing information and images. While providing details, buyers can easily search, browse, and securely pay for the products they want. They can purchase them through various methods. E-commerce systems generally feature user-friendly interfaces and order processing. shopping by offering features such as tracking systems, multiple payment options and customer support. It facilitates the experience. Storage, logistics, and delivery processes are also important parts of this ecosystem. These components ensure that products reach the consumer quickly and reliably. 25 On e-commerce platforms, a single catalog is a platform where all products are listed uniquely and It refers to a structure where each product is defined only once. A single catalog is a catalog where the same product is defined only once. By preventing repeated uploads by different vendors, it prevents duplicate registrations and It aims to improve the user experience. A single catalog allows buyers to see different versions of the same product. It allows sellers to see their prices and conditions on a single page, thus enabling comparison. 30 2 and simplifies the purchasing processes. In the individual catalog, products are categorized according to their unique characteristics. They are grouped and sellers compete by adding price, stock and shipping information for the same product. This system ensures that users have a better shopping experience. In individual catalogs, although the same product is offered by different sellers, the buyer can only purchase a single product. It is very important that the page is seen. This approach prevents duplicate and identical listings of the same product on the platform. By preventing content from appearing on repetitive pages, users can have a more organized and efficient shopping experience. This allows sellers to have a positive experience. They can display their own prices and stock levels on a single product page. They differentiate themselves by specifying the status and shipping options; users can see what different sellers offer. It allows customers to compare prices and service terms on a single page, enabling them to make the best choice. If a unique catalog is not provided, the same product will be repeated many times on the platform. This situation results in unnecessary clutter in the search results for users. This makes it difficult for buyers to access the information they are looking for. Accuracy and suitability of product data in individual catalogs are crucial on e-commerce platforms. Products should be classified into the correct categories and as many features as possible should be added. This situation... It makes it easier for buyers to find and compare the right products. If the data is incorrect, 15 Users are encountering misleading information and may purchase the wrong products. Furthermore, the products... Being placed in the wrong categories causes confusion in search results and negatively impacts user experience. This is negatively affected. The lack of product features also hinders buyers' decision-making. This makes it difficult. The efficiency of individual catalogs, the accuracy and suitability of product data... control, classification in the correct categories, and a maximum number of product features (20) It depends on recording. Existing single catalog solutions automate processes and reduce labor costs. It falls short in that sense. Today, artificial intelligence and machine learning are used in many fields such as health, finance, education, and commerce. By offering revolutionary solutions, these technologies touch every aspect of our lives. Major advantages include speeding up processes, reducing costs, and improving user experience. 25 Artificial intelligence and machine learning techniques enable the improvement of existing single catalog solutions. It has only recently started to be used. However, these techniques are not suitable for existing single catalog solutions. It is not yet able to provide sufficient assistance. Current machine learning and artificial intelligence-based approaches... While solutions play an important role in automating these processes, they are insufficient in many cases. remains. Problems such as incorrect categorization, missing product features, and faulty data matching affect all 30 This negatively impacts user experience and reduces the platform's credibility. Therefore... More advanced algorithms, customized AI models, and data that provides high accuracy. 3 Verification systems are needed. Such a solution both increases user satisfaction and... This also increases the platform's competitiveness. Document number US2004267614A1 describes a customized e-commerce platform based on customer segmentation. The text discusses catalog management and systems, as well as product presentations based on customer segmentation. and to offer personalized shopping experiences through a pricing structure based on commercial agreements. 5 This aims to increase customer satisfaction through the aforementioned personalized catalog structure. It offers a competitive advantage to trading companies. Document US2007150376A1 presents a catalog viewing method and system. The elements are displayed in a static view, and detailed information for the selected element is quickly accessible. The aim is to provide this. The system mentioned allows users to access multiple products simultaneously. 10 Detailed views and comparison view for examination and comparison. It offers. The document numbered US2008255963A1 describes how users can purchase products listed in an e-catalog. visualizes the user's past purchases with different graphic styles, showing them what they have purchased in the past or have yet to buy. A method is described that allows for distinguishing between products that have not been purchased. The relevant document describes a hierarchical structure of 15. Catalog information organized in a tree structure is retrieved from the database via a device. is presented to the user. However, the relevant document mentions individual products used on e-commerce platforms. There is no solution available for their catalogs. In conclusion, the accuracy and suitability of product data can be checked, ensuring that the products are correct. It allows for the classification of products into categories, and the recording of a maximum number of product attributes. A method that provides this is needed. Purpose and Brief Description of the Invention The purpose of the invention is to create a unique product category so that products can be easily found on online shopping platforms. The goal is to ensure the creation of catalogs. Another purpose of the invention is to enable verification of the accuracy of product data. 25 Another purpose of the invention is to ensure that products are assigned to the correct categories. The invention is achieved by entering data into a vendor screen or by receiving data from an integrator via an API. It is a single catalog and category method where products can be easily purchased on online shopping platforms. Creating individual product catalogs for easy identification, and verifying the accuracy of product data. and to ensure that products are assigned to the correct categories; 30 4 a product installation server where the product installation interface data and the product installation API are generated, Comparison results are a system where interface data is generated and product matching processes are performed. the back office server, where an error interface data and an update error interface data are generated and a validation server where products are validated according to a rule set, receives incoming product data. a catalog server where product catalog data is recorded and contains the vendor's inventory, 5 an inventory server and a data collection server where update interface data is generated through its elements, - product installation interface data generated on the aforementioned product installation server After being transmitted to the aforementioned vendor screen, the product data is displayed on the product upload interface. by entering and / or using an API created on the product upload server, the integrator's product 10 transmitting the data and comparing the product data with the back office transmitted to the server, - On the back-office server, new product data and previous product data are displayed according to a threshold value. comparison, - On the back office server, the product data received according to the comparison result is compared with the previous product 15 If there is no match above a certain value between any of the data points, the incoming product data will be replaced with a new one. Before being added to the catalog as a product, it is sent to the validation server for validation. transmission, - On the back-office server, the new product data is compared with the previous product data based on the comparison result. If there is a match above a certain value between any of the data points, the comparison is 20. a comparison created on the back-office server to submit the results for vendor approval. The results are transmitted as interface data to the vendor's screen. - On the seller's screen, the incoming product data is checked and the incoming product data is determined to be "new". Comparison results of data entry meaning "product" or "one of the previous products". Accessing the interface and transmitting the data entered into the interface to the back-office server, 25 - Comparison results after data verification on the back office server. If data entry meaning "new product" is made into the interface, the incoming product data will be a new one. Before being added to the catalog as a product, it is sent to the validation server for validation. transmission, - Comparison results after data verification on the back office server 30 If a data entry meaning "one of the previous products" was made into the interface, the incoming product will be... the data is transmitted to the inventory server to be added to the inventory and on the inventory server Recording product data, - On the validation server, the incoming product data must be subjected to a validation rule set and If there is no error in the product data in question, it will be added to the catalog as a new product. The product data is sent to the catalog server; if there are errors in the new product data, the incoming product data is checked. For this purpose, error interface data generated on the validation server is displayed on the vendor screen. transmission, 5 - On the vendor screen, checks are performed on the product data in the error interface and the error in question is identified. After errors are corrected by adding / removing data in the interface, the errors are displayed. The deleted product data is transmitted to the catalog server to be added to the catalog as a new product. and saving product data to the catalog server, - New product data is transferred from the catalog server to be sold on the online shopping platform. forwarding to the data collection server, - Data collection is required on the data collection server so that new product data can be updated. Transmitting update interface data generated on the server to the vendor screen, - On the seller's screen, new products can be added / removed by performing data updates in the update interface. updating the data and adding the updated product data to the catalog as an updated product 15 Before being added, it must be submitted to the validation server for validation. - On the validation server, the updated product data must be subjected to the validation rule set. And if there are no errors in the updated product data, the updated product data is added to the catalog. Product data to be transmitted to the catalog server for inclusion and sent to the catalog server. saving; if there is an error in the updated product data, checking the updated product data 20 For this purpose, an update error interface data generated on the validation server is sent to the vendor. transmitted to the screen, - On the vendor screen, checks are performed on the update error interface for the updated product. And the update involves adding / removing data in the error interface to resolve errors. After the errors are resolved, the updated product data, with the errors corrected, will be added to the catalog on 25. transmitting the product data to the catalog server and saving the product data on the catalog server. It includes the steps involved in the process. The invention is achieved by entering data into a vendor screen or by receiving data from an integrator via an API. It is a single catalog and category method where products can be easily purchased on online shopping platforms. Creating individual product catalogs to facilitate finding them, and checking the accuracy of product data. 30 and it relates to a method that ensures products are assigned to the correct categories. 6 Brief Description of the Figures Figure 1 shows the elements of the single catalog and category method and the interactions between these elements. A representative view is given. Reference Numbers Seller screen 5 Integrator Product upload server 40 Back office servers 50 Validation servers 60 Catalog servers 10 70 Inventory servers 80 Data collection servers Detailed Description of the Invention On e-commerce platforms, relevant products are listed when a buyer performs a search. It is required. Sellers selling the product must be listed under the product with their own sales amounts, and the listing must be completed in 15 days. Creating a unique catalog is fundamental because it is for different products and the same product is not listed repeatedly. These are the features. A matching process is performed when products are entered into the system to create a unique catalog. This is necessary. The invention allows products to be easily found on online platforms. Creating individual product catalogs, checking the accuracy of product data, and ensuring products are correct. It relates to a method that enables the assignment of categories. 20 The servers mentioned in the specification will be located in other environments (database, different servers, cloud). (environments, etc.) that can receive data or transmit data to those environments using different tools (API, etc.), or refers to computing environments that can record the data they transmit. In other words, the aforementioned Servers are used for storing and sharing data, hosting web pages, running applications, or performing specific tasks. providing a service involves fulfilling client requests and providing them with the necessary data or service. It is either hardware or software. The invention basically consists of a vendor display (10), an integrator (20), a product upload server (30), a back office server (40), a validation server (50), catalog server (60), an inventory server It consists of (70) and a data collection server (80). The aforementioned seller screen (10) displays the interfaces, allows data entry, and facilitates shopping. It refers to a structure that enables the monitoring of processes. Here, the vendor screen (10) is an information processing a device (e.g., computer), a mobile device (e.g., phone), etc., or simply a viewer (e.g., (monitor) can be. Seller screen (10); product loading server (30), back office server (40), 7 validation server (50), catalog server (60), inventory server (70) and data collection server It interacts with (80). The mentioned integrator (20) brings together different systems to facilitate e-commerce operations. This refers to the automation tools that bring about change. These tools include e-commerce platforms and accounting systems. Data transfer between programs, inventory tracking systems, shipping companies, and marketplace management tools 5 These are software programs that provide order management, stock updates, and more. It aims to manage billing and shipping processes through a single platform for sellers. It can sometimes work with integrators (20). In the invention, integrators (20) can work with product loading. API (Application Programming Interface) provided (opened) from the server (30). To add product data (information) to the catalog via the interface, 10 (30) transmits. The integrator (20) transmits with the product loading server (30) and the inventory server (70). They are interacting. The seller requests an update from the aforementioned seller screen (10) and with the help of integrators (20). It can be found. The mentioned product loading server (30) transmits the product loading interface data to the vendor screen (10) 15 and refers to a server on which the product upload API provided to the integrator (20) is created. The aforementioned product loading interface is transmitted to the vendor screen (10) with the help of a tool (e.g. API). The seller can enter data individually or in Excel into the product upload interface displayed on the seller screen (10). It saves product data in bulk using a file. This data is saved to the product upload interface. The data is transmitted to the relevant server, whichever server needs it, via a data transmission tool. The 20 created... Interface data is saved to the memory unit of the product loading server (30). Product loading The server (30) interacts with the vendor screen (10), integrators (20) and back office server (40). The aforementioned back office server (40) processes the product data in the catalog with newly entered data according to a threshold value. The interface where comparison processes between product data take place and the comparison results are presented. It refers to the server where the data is generated. The aforementioned threshold value comparison, the seller's 25 or the product into which the integrator (20) entered data and which was previously recorded on the back office server (40) It refers to the process of measuring / calculating the proximity of the vectors of the products to each other. The comparison result shows that the products previously saved on the back office server (40), in other words products in the catalog with products entered by the seller or integrator (20) a certain threshold value If a match is found (in proximity), the newly entered data indicates that the product is considered a potential match. 30 Back office server (40); vendor screen (10), product upload server (30), validation server (50) and interacts with the inventory server (70). 8 The back office server (40) includes an artificial neural network. In the preferred configuration of the invention, the back office server (40) includes an artificial neural network. CNN artificial neural network is used in the office server (40). CNN is a subfield of Machine Learning. It uses multilayer artificial neural networks and automatically extracts features from the data. It is learning. The CNN model (40) on the back office server handles a large amount of text (specifically e-commerce). (data) is an open-source model trained with and will use this model to identify products. Vectors are generated. Here, in addition to storing vectors for all products in the catalog, new ones are also created. Vectors are also generated for the input products. The aforementioned CNN model supports 16 different languages. It is an artificial neural network. The data for this model includes the model's algorithm data, cleaning, and preprocessing algorithms. data, generated vector data, current model data, generated interface data, all in the catalog Product data, data for each newly added product, hierarchical data, and the 10 categories of products that have data. Required attribute data, prohibited word data, VAT data for categories, installment data, volumetric weight (desi) data and metadata related to e-commerce and unique catalog processes on the back office server. (40) are saved to the memory unit. In other words, product categories, category tree, product types Mandatory and optional product features and their possible values, installment rules based on product type, product Type-based VAT values ​​are stored hierarchically on the back office server (40) and can be accessed when needed. It is served (transmitted) to other servers. When a product is entered, the correct category of the product is selected. The number of installments and the VAT value are determined by algorithms and models (40) on the back office server. Some examples of data held on the back office server (40) are given below; Category Tree Example: XYZ.com>>Fashion Guru>>Women>>Clothing>>Vest 20 XYZ>>Fashion Guru>>Men>>Clothing>>Vest Example Product Specifications (for a vest) a. Pattern (optional) b. Season (optional) 25 c. Gender (required) d. Color (mandatory, variant feature) e. Body (mandatory, warrant feature) Examples of possible values ​​for a vest's size attribute: a. S 30 b. M c. L d. XL 9 e. XXL Example of Installment Payment Rules: For vests: 12 installments Example of a VAT rule For vest product type: 10% 5 The aforementioned validation server (50) is where product data is checked and an error interface data is generated. and an update error interface data is generated, product data is validated according to a rule set. It refers to the server where validation processes are carried out. On the validation server (50), validation processes are easily carried out. It includes three different machine learning models for both implementation and automation. 10 In the preferred configuration of the invention, the validation server includes model 1, model 2 and EfficientNet (50). The following models are used: Validation server (50); vendor screen (10), back office server (40) and It interacts with the catalog server (60). The aforementioned catalog server (60) refers to the server where all product data is recorded and stored. The catalog server (60) contains a catalog in which all the features of the products are listed. 15 In the catalog server (60), it is necessary to extract product specifications not provided by the vendor and Machine learning models are used to create images and text for cataloging purposes. Product feature extraction is performed with the help of and text-based methods. Catalog server (60); The vendor screen (10) interacts with the validation server (50) and the data collection server (80). The catalog server (60) uses a machine learning model to suggest categories. 20 In the preferred configuration of the invention, CBOW aims to propose categories in the catalog server (60). The model is being used. Category suggestion is deliberately used by sellers for fraudulent purposes. or to help products that were unintentionally opened in the wrong category to be placed in the correct category. This model uses CBOW for the multi-class text classification problem with an unbalanced dataset. The solution is generated using the algorithm. 25 The catalog server (60) uses a machine learning model to extract product features from the image. It is used. In the preferred configuration of the invention, the product is shown in the catalog server (60) with images. The YOLOv8 model is used to extract the feature. Product features are extracted from the images. The YOLOv8-based object recognition model is trained and used on online fashion sales data, and The identified products are categorized according to their characteristics using a stepped model structure and accurate category breakdowns. Classification is provided. Extracting product attributes from product name, product description etc. on the catalog server (60). A machine learning model is used for this purpose. In the preferred configuration of the invention... In order to extract product features from product name, product description etc. on the catalog server (60). A Large Language Model (LLM) model is being used, preferably ChatGPT. It is used. For example, the LLM solution 5 where the hard drive with code HD13500 has a capacity of 500GB. It can be determined. The catalog server (60) also uses text-based methods to extract product features. It includes. In the preferred structure of the invention, the catalog server includes (60) product name, product REGEX (Regular Expression) is used to extract product features from descriptions and other similar characteristics. Expression) methods are used. For example, if the product name includes "white" but the color attribute is blank, this is 10 This can be completed using a REGEX solution. Using machine learning and text-based methods mentioned in the catalog server (60) This allows sellers to easily find products according to their desired specifications. Some products and variants... The properties must have standard values ​​defined on the back office server (40). For example, “vest” The product must have a size specification of one of the following values: S, M, L, XL, XXL. If the aforementioned 15 If it does not have one of the defined values, the text described above and used for feature extraction will be used. Machine learning models based on methods can extract the most from the entered name, features, and images of the product. It adds the appropriate value to the product attributes. For example, if the size attribute is entered as 36, and the product name is... If the word "Small" appears or if the body information in the images is "S", these are text and image processing capabilities. It is determined by learning models and added to product features as an S feature value. More than 20 clearly in the catalog server (60), product data received from the vendor or integrator (20) If there are missing and / or enrichable features, REGEX solution, LLM solution, CBOW solution and It is identified by finding the solution with YOLOv8. The products are presented in the catalog with all the information about the main product and its variants. Catalog data. 25 for use in product listing / detail / filtering / search applications on web / mobile applications. Data transmission is served to the relevant servers via a messaging structure (e.g., Apache Kafka). Product Having many features and, as much as possible, having defined standard values ​​for product search and Filtering helps the buyer easily find the product, while also ensuring brand differentiation. For SEO (Search Engine Optimization), traffic needs to be managed correctly. It contributes to 30 product categories, product names, and features using machine learning methods. The product quality score is determined based on existing product features when a product is added to the catalog. This score is determined and used as an input when listing search results, and it identifies the seller and the product. 11 It encourages the enhancement of its features. Below is the JSON sent to the catalog server (60). A sample message is provided. { "_id" : ObjectId("57690d11c0140edcb8731944"), "createdAt" : ISODate("2021-03-05T09:04:08.967+0000"), 5 "createdBy" : "user-0", "fields" : { "Product Description" : { "value" : "MADE FROM FIRST CLASS POLYESTER FABRIC. CURTAIN ROD MOUNTING ACCESSORIES INCLUDED." INCLUDED. There is a 4 cm allowance in roller blinds. For example, if you purchase a 60 cm wide product, the net width of the blind will be 10 cm. Its width is 56 cm. "mandatory" : false }, "hbSku" : { "value" : "EAKSODMTP10398", 15 "mandatory" : false }, "TaxVatRate" : { "value" : "8", "mandatory" : false 20 }, "Currency" : { "value" : "TL", "mandatory" : false }, 25 "Image1" : { "value" : "petrolmavi.jpg", "mandatory" : false }, "VariantID" : { 30 "value" : "TPED122", "mandatory" : false }, "ProductName" : { "value" : "T*** Polyester Roller Blind P-122 Petrol Blue Skirted 80x200", 35 "mandatory" : false }, "StockQuantity" : { "value" : "25", "mandatory" : false 40 }, "Barcode" : { 12 "value" : "3425589083680", "mandatory" : false }, "Kg" : { "value" : "5", 5 "mandatory" : false }, "WarrantyPeriod" : { "value" : "24", "mandatory" : false 10 }, "Brand" : { "value" : "T***", "mandatory" : false }, 15 "Theme" : { "value" : "Flat", "mandatory" : false }, "FabricTour" : { 20 "value" : "Polyester", "mandatory" : false }, "CurtainFeatures" : { "value" : "Polyester", 25 "mandatory" : false }, "Colour" : { "value" : "Petrol blue", "mandatory" : false 30 }, "Fabric Weaving" : { "value" : "Polyester", "mandatory" : false }, 35 "Package Included" : { "value" : "Curtain, curtain rod, invoice", "mandatory" : false }, "Dimensions": { 40 "value" : "80X200", "mandatory" : false }, 13 "merchantSku" : { "value" : "EAKSODMTP10398", "mandatory" : false } }, 5 "retail" : false, "listingStatusList" : [ ], "isMatched" : true, "merchant" : "user-3", 10 "merchantSku" : "EAKSODMTP10398", "status" : "MATCHED", "validationStatus" : "FAILED", "productType" : { "name" : "Curtains", 15 "productTypeId" : NumberInt(997), "buyingCategoryId" : NumberInt(80794182), "buyingCategoryName" : "Curtain", "parentBuyingCategoryId" : NumberInt(80089022), "parentBuyingCategoryName" : "Home Textile Store", 20 "deleted" : false }, "uploadDate" : ISODate("2016-06-21T09:46:00.000+0000"), } Below is an example of a JSON message sent from the product catalog to the data collection server (80). { "Id": "HBC000060ZS1T", "ProductId": "HBC000060ZS1T", "Name": "P*** S*** G*** A50 Cover Nostalgia Pattern Design Transparent Super Silicone Case", 30 "Brand": "P***", "CatalogName": "Phone", "DefinitionId": "529", "DefinitionName": "Phone Cases", "BuyingProductType": "6804", 35 "InstallationRate": 12, "BuyingCategoryId": "80794416", "Variants": [ { "Id": "HBCV000060ZS1U", 40 "Name": "Printify Samsung Galaxy A50 Cover Nostalgia Pattern Design Transparent Super Silicone Case" "TaxVatRate": "20", 14 Freight: 1.00 "WinnerFreight": "1.00", "Barcodes": [ "prtsprslknsmA50043d" ], 5 "Media": [ { "ImageFileName": "110000643300800.jpg", "LinkFormat": "https: / / productimages.hepsiburada.net / s / 777 / {size} / 110000643300800.jpg", 10 }], "Video": [], Warranty Period: 6 "VariantClassifications": [ { 15 ClassId: "529", "ClassName": "Phone Cases", "AttributeId": "phone_model", "DisplayValue": "G*** A50", "Name": "Phone Model", 20 "Value": "0000QPGM" }, { ClassId: "529", "ClassName": "Phone Cases", 25 "AttributeId": "compatible_brand", "DisplayValue": "S***", "Name": "Compatible Brand", "Value": "4316" }, 30 { ClassId: "529", "ClassName": "Phone Cases", "AttributeId": "case_type", "DisplayValue": "Back Cover", 35 "Name": "Case Type", "Value": "14065" }, { "ClassId": "529", 40 "ClassName": "Phone Cases", "AttributeId": "material_type", "DisplayValue": "Silicone", "Name": "Material Type", "Value": "9077" 45 }, ], "CreateDate": "2024-03-02T21:11:19.8600000" } } The mentioned inventory server (70) is the server through which vendors can track product names and stocks. It states that product name data and stock data are stored in the memory unit of the inventory server (70). The inventory server (70) records itself for each product added to and removed from the catalog. It is updating. The inventory server (70) contains the inventory that vendors can see. Vendor 5 If a change is made to the product in the inventory, the data is transmitted to the data collection server (80). The vendor screen (10) and the integrator (20) interact with the inventory server (70). Vendors The inventory can be accessed directly or through integrators (20). Sellers can access it directly or through integrators. (20) They can create an update request regarding the products in their inventory. Inventory On the server (70), thanks to the interfaces created, inventory information can be tracked on the vendor screen (10). It can be done. Inventory server (70); vendor screen (10), integrator (20) and back office server It interacts with (40). The mentioned data collection server (80) is located on the back office server (40) and the catalog server (60) The product data found differs from other data (price, campaign, etc.) that are not present on those servers. It collects and combines data from various sources (websites, devices, APIs) and presents the combined data, resulting in a 15 This refers to a server where the update interface data is generated. Data collection server. (80) provides a central perspective by receiving data from distributed sources and especially data management It facilitates their processes. Data collection server (80); vendor screen (10) and catalog server It interacts with (60). Before the working method of the invention is described in detail, some definitions will be made. 20 The interface creation process to be described refers to a series of traditional interface creation processes. The interface is designed using a design tool (Figma, Adobe, etc.) to suit the needs. It is being designed (drawn, visualized) and details such as user interactions and interface layout are being considered. is determined. The designed interface is created through a series of front-end development processes (HTML, CSS, The interface is coded using languages ​​like JavaScript, React, Swift, etc. The coded interface is then presented as a 25-inch screen. It goes through the backend development process (RESTful API, GraphQL, etc.) and the interface. The elements (buttons, etc.) will become functional (e.g., receiving / transmitting data from a server) The data to be displayed on the interface is coded. During the backend development process, the data is developed in the frontend. It is processed in a way that will suit the party. In other words, in the back-end-front-end communication process. Data received from or transmitted to the server must conform to specific formats (JSON, XML). 30 It is processed in the backend process. Finally, the interface code data is displayed to a browser (web-based interface, etc.). or converts it into visual data using a graphics engine (SurfaceFlinger, UIKit) (rendering) (processing). The calculations required for visual-graphic processing are performed by the CPU (Central Processing Unit). This is done via the GPU (Graphics Processing Unit) or the PC (Graphics Processing Unit). For example 16 While GPUs are used for drawing visuals and animations, CPUs handle data processing and business logic. Operations related to the processes can be performed. Interface data converted into visual data. The information is transmitted to the viewer (monitor, television, computer, phone, etc.) via a device. Here The tools refer to different hardware or software elements (API, HDMI, GPU, etc.). In the invention... For vendor screen (10) interactions, which represent a viewer, the API tool is used. 5 The invention involves a series of process steps in its operation. First, the aforementioned product loading... A product loading interface data generated on the server (30) is a tool on the mentioned vendor screen (10). It is transmitted with the help of. Subsequently, the product loading interface is displayed on the seller's screen (10). Sellers enter product data. Here, sellers individually enter product data into the product upload interface. While it can access the product through the application, it can also enter data into the product upload interface in Excel format. 10 with integrators (20) employees transmit product data to integrators (20) and product data entry is done by integrators (20) In other words, the integrator (20) creates an API on the product installation server (30). It transmits product data with the help of the product upload interface (10) on the seller screen and Product data transmitted from integrators (20) to the product loading server (30) for comparison purposes from the product installation server (30) to the back office server (40) with a data transfer tool (Rsync, Apache 15 (Kafka, etc.) is transmitted. The invention describes the data used for transmitting and receiving data between servers. The communication medium is Apache Kafka, and the data type is JSON. On the back office server (40), new incoming product data and previous product data are compared according to a threshold value. They are compared. The matching process involves product name, product specifications, product brand, and product images. using the data, a CNN (Convolutional Neural Network) on the back office server (40) It is done with the Neural Network model. On the back office server (40), the matching process is performed using the aforementioned CNN. This is done by comparing the newly entered product data with other product data in the catalog. On the back office server (40), in the first stage, a traditional cleaning algorithm is used to clean the catalog. Removing special characters from registered product data, converting all characters to lowercase. Some preliminary processing is performed. The product data that has undergone pre-processing is then used as input to the CNN model. Vector data of these data (texts) are produced by giving them. On the back office server (40), CNN Vector data generated by the model can be entered into a search engine (Milvus, Pinecone, Elasticsearch, etc.). The search vector data is being added (saved) and fed into the CNN model to find the search. Products with the closest characteristics to the product are separated. On the back office server (40), the CNN model To improve precision and recall performance, 30 steps are added after the search results. Additional checks such as barcodes and product specifications are being added. On the back office server (40), when a new product is entered into the product loading server (30), the entry is made. Product data is transmitted from the product loading server (30) to the back office server (40) and CNN 17 With the help of this, the product vector is extracted from the product's feature data, previously on the back-office server. (40) subject to vector search with vectors of products stored in memory unit It is being held. According to vector similarity search, new product data is above a certain threshold value. It is matched with the product that has the closest matching vector. As a result of the matching, if the product If found in the catalog data, the seller can create a product listing with their own price and stock information. 5 It is expected. If not found, it will be subject to validation to be added as a new product. is being held. Traditional methods of product matching using barcodes rely solely on vendor data, and This leads to duplicate products being included in the catalog. However, in our invention, the back office... Thanks to these operations performed on the server (40), machine learning and vector search can be done in just 10 Instead of barcodes, a much more consistent result is obtained by using many features such as product name, specifications, and images. Matching is performed and the unique catalog is secured. On the back office server (40), the comparison result is based on a predetermined threshold value. According to the data, there must be a match above a certain value between the incoming product data and one of the previous product data points. Otherwise, the incoming product data must be validated before being added to the catalog as a new product. 15 It is transmitted to the validation server (50) via the transmission medium. On the back office server (40), comparison According to the result, there is a specific value difference between the newly received product data and one of the previous product data. If a match is found, the comparison results data is submitted to the back office for vendor approval. Comparison results generated on the server (40) are displayed on the vendor screen (10) via interface data. is being transmitted. 20 On the seller's screen (10), the product data received in the comparison results interface is checked by the seller. The product data received is processed and the seller marks it as a "new product" or "one of the previous products". This means that data input is entered into the comparison results interface, and the data entered into the interface is then processed in the background. to the office server (40) for “new product” or “one of the previous products” check with previous products It is transmitted with the help of a tool for comparison. 25 On the back office server (40), after checking the data, the comparison results interface If a data entry indicating "new product" was made, the incoming product data will be added to the catalog as a new product. before being added, it is sent to the validation server (50) via a data transmission tool to be validated. It is being transmitted. On the back office server (40), after checking the data, the comparison results interface is 30 If a data entry is made indicating "one of the previous products," the incoming product data is added to the inventory. 18 Product data is sent to the inventory server (70) to be added and is on the inventory server (70) It is being recorded. On the validation server (50), product data received from the back office server (40) is validated. It is subjected to a validation rule set, and if there is no error in the product data in question, a new product is released. If there is an error in the new product data being sent to the catalog server (60) to be added to the catalog, 5 An error interface is created on the validation server (50) to check the incoming product data. The data is transmitted to the vendor screen (10) with the help of a tool. The rule set algorithm mentioned here The product fields must not be empty, the product must be unique, and the entered product data must not contain any prohibited words. It includes controls such as the absence of (Table 1). Table 1: Non-visual validation controls (rule set) 10 Areas Rules The barcode must be unique and comply with the brand's barcode standards. The main product name must not be left blank and must not contain prohibited words. The brand must be defined in the system. Product descriptions must not contain prohibited words. Category: One of the available categories must be selected. Product Type: One of the available product types must be selected. Variant Type: If there is a variant group, the type cannot be empty. The VAT field should not be left blank. The volumetric weight (Desi) cannot be less than 0. The warranty cannot be less than 0. Properties: The required property value cannot be empty. Visual data checks are also performed on the validation server (50). Here, the known technique In this case, manual controls can be used to check +18 images containing child models, solely for this purpose. This is done quickly using a machine learning method trained to perform the control. 19 On the validation server (50), different models are trained to perform visual validation. The models mentioned are the EfficientNet model, and in addition to this model, Model 1 and Model 2. Validation processes are carried out using the models mentioned. On the validation server (50), we can call it the basic model and easy tasks Model 1 is being trained to do this. Model 1 is used as a basic feature extractor (5). It is used and is based on the EfficientNet architecture. A ready-made model directly for Model 1. It is unused and customized for e-commerce and individual catalogs. Model 1 is for validation processes. Depending on the required features, some layers can be removed from the neural network and replaced with a new one. The classifier is trained using a fully connected neural network. On the validation server (50), we can call it an advanced model and it has 10 sensitive tasks. Model 2 is being trained to perform more complex or delicate tasks. In addition to the feature extractor EfficientNet model, a specially developed and convolutional artificial neural network has been used. Convolutional neural network (CNN) features were used. Two extracted features are noteworthy. It is combined with the attention mechanism. The resulting feature is fully connected. connected to the neural network and, depending on the module's requirements, undergo regression or classification. 15 It has been trained with (classification) objectives. In the validation server (50), the trained models will be used in the inference phase. Parameters designed to meet the requirements of high response speed and high classification performance. Models are designed taking into account the number of data sets. When designing the architecture, a data set is considered. The feature outputs of the EfficientNet model trained with and sub-model 1, which produces a separate feature, are shown in the diagram. The outputs are used by combining the focus (attention) approach. With this approach, the images... The aim is for the model to understand which areas need to be focused on. This allows for visual... When resolution quality is investigated, errors in the details of the images are detected with higher accuracy. The preferred configuration of the invention involves the EfficientNet model with the ImageNet dataset. They are being trained. 25 On the validation server (50), visual checks are performed for the headings given in table 2. For this purpose, the images (visual data) of the products available in the catalog are initially labeled using manual marking. Then, machine learning models are trained with these images. Afterwards, what the model does... Manual feedback is being added to the labels, and the models continue to be trained at regular intervals. This is done. Then, visual checks are performed with the results of the models. 30 Table 2: Visual verification checks Control Rules Low resolution. The image is of low resolution. The image should be between 250 pixels and 250 pixels. Child models. Inclusion of child models. Text. The image should contain text. The logo should be in the image. Watermark: The presence of a watermark in the image. Assorted: Products that come in more than one color. Credit card. Credit card / cash. +18 The image contains adult content. When entering product data (seller screen (10) or integration), the main product and its associated different features are entered. These are added as variants. For example, the XYZ brand vest product comes in sizes S, M, L, XL, and XXL. Depending on its characteristics and color options (white, black, gray), it can be found in different variants. The situation is that when looking at a product, other variants of the same product can also be easily found. This approach ensures that product features are validated once and variantly on the validation server (50). The features are subject to validation for each variant. These validations are carried out for Model 1, Model 2, and Model 3. This is done using model 2 and EfficientNet. 10 Product data checks are performed by the seller on the error interface displayed on the seller screen (10). This is done by the vendor adding / removing data in the error interface in question to correct errors. After the errors are corrected, the product data with the corrected errors is added to the catalog as a new product. Product data is transmitted to the server (60) and recorded in the catalog server (60). New product data 15 from catalog server (60) to data collection server to be sold on online shopping platform (80) is transmitted by a data transmission medium. In order to update the new product data on the data collection server (80), (80) The generated update interface data is transmitted to the vendor screen (10) via a tool. On the vendor screen (10), by adding / removing vendor data in the displayed update interface The new product data is being updated, and the updated product data is being added to the catalog as an updated product. 21 before being added, it is sent to the validation server via a data transmission aid to be validated (50) It is transmitted. On the validation server (50), the updated product data is subject to the validation rule set. The updated product data is kept and, if there are no errors in the updated product data, it is added to the catalog. The data is transmitted to the catalog server (60) with the help of a data transmission tool to be added and the catalog Product data is saved to the server (60). On the validation server (50), the updated product data is 5 It is subject to a validation rule set, and if there are errors in the updated product data in question... an update created on the validation server (50) to check the updated product data Error interface data is transmitted to the vendor screen (10) via a tool. On the vendor screen (10) In the displayed update error interface, the vendor is performing checks on the updated product and... The topic is about fixing errors by adding / removing data in the update error interface. 10 Subsequently, the updated product data, with the errors corrected, is sent to the catalog server to be added to the catalog. (60) is transmitted with the help of a vehicle and product data is recorded on the catalog server (60). Our invention enables the creation of a single product catalog using machine learning methods. It ensures that products are assigned to the correct categories, and verifies the accuracy of product data. Checking whether it is suitable for sale, product information not sent by the seller 15 By filling it in, it creates a rich catalog and in all these aspects differs from the previous technique. 25

Claims

22 REQUESTS 1. The invention is based on data input to a vendor screen (10) or data from an integrator (20) via an API. It is a single catalog and category method that can be obtained with the help of; its feature is online shopping. Creating individual product catalogs so that products can be easily found on platforms, product 5. Checking the accuracy of the data and assigning products to the correct categories to provide; a product installation where product installation interface data and product installation API are generated server (30), comparison results interface data is generated and product matching a back office server (40) where the operations are performed, an error interface data and an update A 10-point system where error interface data is generated and products are validated according to a rule set. validation server (50), where incoming product data is recorded and product catalog data is included a catalog server (60), an inventory server (70) containing the vendor's inventory and a elements of a data collection server (80) where update interface data is generated by,  a product installation interface created on the mentioned product installation server (30) 15 After the data is transmitted to the aforementioned vendor screen (10), product loading Product data entry in the interface and / or product upload server (30) With the help of an API created, the integrator (20) transmits the product data and the said Product data is sent to the back office server (40) for comparison purposes.  On the back office server (40), new incoming product data and previous product data are placed at a threshold of 20 Comparison based on value,  On the back office server (40), the product data received according to the comparison result is compared with the previous If there is no match above a certain value between any of the product data, the incoming product will be rejected. data to be validated before being added to the catalog as a new product 25  on the back office server (40), with the new product data according to the comparison result if there is a match above a certain value between one of the previous product data on the back office server (40) to submit comparison results for vendor approval Transmission of the generated comparison results interface data to the vendor screen (10),  On the seller screen (10), the incoming product data is checked and the said incoming product is 30 data entry meaning "new product" or "one of the previous products" Entering the comparison results interface and processing the entered data into the back office to be sent to the server (40), 23  Comparison after checking the data on the back office server (40). If data entry meaning "new product" was made into the results interface, the resulting product will be displayed. data to be validated before being added to the catalog as a new product to be sent to the validation server (50),  Comparison after checking the data on the back office server (40) 5 a data entry into the results interface meaning "one of the previous products" If done, the incoming product data is sent to the inventory server (70) to be added to the inventory. Transmission and recording of product data on the inventory server (70),  On the validation server (50), the incoming product data is subjected to a validation rule set. and if there is no error in the product data in question, it will be added to the catalog as a new product. forwarded to the catalog server (60) to be added; if there is an error in the new product data a validation server (50) is created to check the incoming product data Error interface data is transmitted to the vendor screen (10),  On the seller screen (10), checks are made on the product data in the error interface and the word The topic is resolving errors by adding / removing data in the error interface. 15 Subsequently, the product data with the errors corrected is added to the catalog as a new product. transmitting to the catalog server (60) and product data to the catalog server (60) recording,  New product data is transferred from the catalog server to be sold on the online shopping platform. (60) transmission to the data collection server (80), 20  Data collection is required on the data collection server (80) so that new product data can be updated. an update interface data generated on the server (80) to the vendor screen (10) transmission,  On the vendor screen (10), new data can be added / removed from the update interface. Updating product data and updating product data into an updated product 25 before being added to the catalog, it must be sent to the validation server (50) for validation. transmission,  On the validation server (50), the updated product data is subject to the validation rule set. and if there are no errors in the updated product data in question, the updated product the data is transmitted to the catalog server (60) to be added to the catalog and catalog 30 Saving product data to the server (60); if there is an error in the updated product data Checking the updated product data on the validation server (50) Transmission of generated update error interface data to the vendor screen (10),  On the seller screen (10), checks of the updated product in the update error interface The process involves adding / removing data in the update error interface. 35 24 After the errors were corrected, the updated product with the errors fixed. the data is sent to the catalog server (60) to be added to the catalog and the catalog saving product data on the server (60) It is characterized by including process steps. 10 20

Images