Multi-language intelligent product comparison and editable quotation system

Abstract

The invention relates to the technical field of foreign trade information processing, and provides a multi-language intelligent product comparison and editable quotation system. The system comprises a data acquisition module, a product comparison module, a quotation generation module, an editable interface module and an output and sending module. Compared with a traditional manual process, the system remarkably improves the quotation efficiency, reduces the error rate, and meets the multi-language business requirements of transnational enterprises.

Description

Technical Field

[0001] The present application relates to the fields of computer information processing, natural language processing and office automation, and in particular to a multilingual intelligent product comparison and editable quotation system. Background Art

[0002] With the growing demand for international trade and cross-border offices, foreign trade companies are currently facing multiple challenges in product promotion and customer quotations: 1. Inconsistent multilingual data – Sources such as Chinese and foreign product pages, Excel files, and API interfaces come in diverse formats, making standardized cleaning and integration difficult. The templates for compiling multilingual matching quotes are fixed and cannot be flexibly customized.

[0003] 2. Inefficient semantic comparison – Traditional rule matching or manual translation comparison is time-consuming and error-prone; data acquisition requires manual copying and pasting or logging into multiple platforms.

[0004] 3. Rigid quotation generation—The existing system only supported single-language or static templates, with no support for online editing or dynamic field customization. The batch sending process was cumbersome, making receipt tracking difficult. System scalability was poor in high-concurrency and multi-tenant scenarios.

[0005] 4. Poor data real-time performance – Existing technologies lack a one-stop, automated, multilingual, editable, and batch-sending intelligent quotation solution. Manual processing is unable to meet the real-time quotation needs of e-commerce platforms or large clients. Therefore, a system that can automatically acquire, intelligently compare, and generate editable multilingual quotations is urgently needed to improve efficiency and accuracy. Summary of the Invention

[0006] The purpose of this application is to provide a multilingual intelligent product comparison and editable quotation system to solve one or more of the above-mentioned technical problems. A multilingual intelligent product comparison and editable quotation system includes: a data acquisition module, a product comparison module, a quotation generation module, an editable interface module, and an output and sending module; wherein: A data acquisition module, configured to acquire product data of a plurality of products, wherein the product data includes product names, descriptions, and images; The product comparison module is used to: extract features from product data to obtain product vectors; calculate the similarity between product vectors; perform clustering and deduplication based on similarity thresholds and regional weights, and output the top-K most matching product pairs; A quotation generation module is configured to fill in the quotation template with the output of the product comparison module to generate a quotation document; the filled-in information includes the product name, images of our / target products, bilingual descriptions, FOB price, MOQ, and remarks; The editable interface module is used to: receive modification instructions for the quotation document through a visual form interface to obtain the final quotation document; The output and sending module is used to export the final quotation file into a target format file and send it to a target address.

[0007] Optionally, it also includes: The operation and security module is used to: deploy with Terraform / IaC and Docker / Kubernetes management systems; maintain asynchronous task queues and retry strategies through RabbitMQ / Bull; connect to Prometheus / Grafana to monitor service health and queue lengths, and configure alarms; enable Istio ServiceMesh to implement mTLS, circuit breaking, and traffic management; use JWT+RBAC and audit logs to ensure multi-tenant data isolation and compliance. Optionally, the data acquisition module includes: a website crawling submodule, a keyword crawling submodule, a file uploading submodule and a data cleaning submodule; wherein, The website crawling submodule performs headless browser deep rendering and DOM extraction based on Puppeteer or Playwright; The keyword crawling submodule is used to extract keywords based on the URLs and files input by the user; The file upload submodule is used for streaming parsing of Excel / CSV and image text recognition using TesseractOCR; The data cleaning submodule cleans, removes duplications, and standardizes the original product data, and stores it in a structured manner.

[0008] Optionally, the product comparison module maps multilingual texts into vectors and stores them in a vector database through automatic translation services and BERT or Sentence-Transformers embedding models.

[0009] Optionally, the product comparison module further includes an image feature extraction submodule based on ResNet or EfficientNet, and fuses and compares the text vector with the visual vector.

[0010] Optionally, the quotation generation module loads a dynamic template based on exceljs or docx-template, and automatically fills in fields, injects a logo, and beautifies the format.

[0011] Optionally, the editable interface module is used for online inline table editing, Undo / Redo, version snapshot and field locking functions.

[0012] Optionally, the output and sending module exports via Excel / PDF / API and sends in batches to a target address via SMTP or RESTful.

[0013] Optionally, the operation and security module includes Terraform IaC, Docker / Kubernetes deployment, RabbitMQ / Bull task queue and Prometheus / Grafana monitoring.

[0014] Optionally, the system is deployed in a cloud environment and uses MongoDB sharding or Shard Key and JWT+RBAC to achieve multi-tenant isolation and permission control.

[0015] Optionally, the system further includes an online editing module; The line editing module is configured to support Flutter or React Native mobile app breakpoint resumption and offline caching functions.

[0016] Optionally, the operation and security module also includes integrated WAF and DDoS protection strategies.

[0017] Optionally, a Webhook notification submodule is also included to automatically push the download link and status to the external system after the quotation is generated.

[0018] Optionally, an unmatched item report submodule is included for automatically generating and exporting a detailed list of unmatched products of the buyer or seller.

[0019] Optionally, it also includes: A main account and multiple sub-account management subsystems are used to share the "read" and "sent" status of overseas customers; When any sub-account performs a view or email sending operation on a target customer, the system will synchronously update the status flag in the customer views of all sub-accounts.

[0020] Optionally, the quotation generation module is further configured to: Automatically add a clickable hyperlink to each product information field in the quotation document, which points to the product's detail page on the seller's or buyer's company website.

[0021] This application has the following beneficial effects: 1. High degree of automation: Full-process automated capture and matching completely replaces manual data collection, saving more than 90% of labor costs; 2. Accurate cross-language matching: Combining automatic translation with a deep learning embedding model, it supports deep learning comparison in 15+ languages ​​with an error rate of <5%.

[0022] 3. Flexible Editability: Dynamically editable templates meet corporate branding needs. The quotation template engine and editable interface module support online field adjustment, logo insertion, and header / footer insertion to meet corporate personalized needs. 4. Multi-format output: Multiple formats, one-click export, and batch sending improve delivery efficiency. Generate Excel, PDF, API, and other formats at once, facilitating system integration and batch sending. 5. Strong scalability: Microservices + containerized architecture enables smooth and elastic scaling under high concurrency. The layered modular architecture and asynchronous task queue design support smooth scaling in high-concurrency scenarios and adapt to massive data processing; 6. Easy to deploy and maintain: Multi-tenant isolation and full-link monitoring ensure secure and stable operation. Support for containerized deployment, Terraform infrastructure as code, and Prometheus / Grafana full-link monitoring ensures low operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings illustrate exemplary embodiments and constitute a part of the specification. Together with the description of the specification, they serve to explain exemplary implementation of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals designate similar, but not necessarily identical, elements.

[0024] Figure 1 Schematically illustrates a schematic diagram of the hierarchical modular architecture of a multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 2 The following schematically shows a functional block diagram of a multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 3 Schematically shows a data acquisition and comparison process diagram of the multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 4 Schematically shows a schematic diagram of a quotation generation and editing interface of a multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 5 Schematically shows a schematic diagram of the output and sending module of the multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 6 Schematically shows a schematic diagram of the operation and security module of the multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 7Schematically shows a batch sending diagram of the multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 8 A flowchart of a keyword rapid quotation system for a multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application is schematically shown; Figure 9 A flowchart of the whole table translation of the multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application is schematically shown; Figure 10 Schematically shows a client system architecture diagram of a multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application; Figure 11 The computer hardware architecture diagram of the multilingual intelligent product comparison and editable quotation system according to the first embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of this application more clear, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0026] It should be noted that the descriptions of "first", "second", etc. in the embodiments of the present application are for descriptive purposes only and should not be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0027] In the description of this application, it should be understood that the numerical labels before the steps do not indicate the order in which the steps are executed. They are only used to facilitate the description of this application and to distinguish each step. Therefore, they cannot be understood as limitations on this application.

[0028] In this specification, unless otherwise specified, the following terms have the following meanings: Product matching: This refers to extracting features from multimodal information such as product names, descriptions, and images from different sources (web pages, files) through natural language processing and vector retrieval technology, and calculating similarity based on cosine similarity or other distance metrics to determine the matching relationship between the target product and our products.

[0029] Editable templates: These templates are used in the quote generation module to dynamically populate quote data based on pre-defined field layouts, styles, and company branding elements, using Excel. They allow users to modify key fields (such as price, MOQ, and notes) online or offline, and are synchronized to the exported file in real time.

[0030] Asynchronous task queue: refers to the message queue architecture implemented based on RabbitMQ or Bull, which is used for scheduling and concurrent execution of time-consuming tasks such as crawling, OCR / translation, semantic comparison, document generation, etc., to ensure the stability and scalability of the system in high-concurrency scenarios.

[0031] like Figure 1 As shown, this embodiment provides a multilingual intelligent product comparison and editable quotation system, including: a data acquisition module, a product comparison module, a quotation generation module, an editable interface module, an output and sending module, and an operation and security module; wherein: The data acquisition module is used to acquire product data of multiple products, wherein the product data includes product name, description and image. Specifically, Figure 2 As shown, the data acquisition module can be used to receive URLs, files or keywords input by users; automatically call crawler services, OCR engines and third-party interfaces to capture multi-source product data; clean, deduplicate and standardize the original product data; and store the structured results in the data storage system.

[0032] The product comparison module is used to: extract features from product data to obtain product vectors; calculate the similarity between each product vector; perform clustering and deduplication based on the similarity threshold and regional weight, and output the most matching Top-K product pairs. Specifically, Figure 2 As shown in the figure, the product comparison module can perform word segmentation, embedding vectorization, and visual feature extraction on the cleaned and multilingual product names, descriptions, and images; calculate the similarity between each product vector; perform clustering and deduplication based on the similarity threshold and regional weight, and output the most matching Top-K product pairs.

[0033] In some embodiments, as Figure 3As shown, first, the user enters the URL or keyword; then, based on the data acquisition module, data is obtained by crawling web pages or calling APIs, and Excel, CSV format files and files recognized by OCR technology are parsed, and the obtained data is cleaned and standardized; then, based on the product comparison module, the text is pre-processed by word segmentation, translation, etc., and then converted into Embedding vectors, similarity calculations are performed, and clustering and deduplication are performed, thereby completing the data acquisition and comparison process for the entire core business.

[0034] Furthermore, the product comparison module can perform multilingual text preprocessing (automatic translation, word segmentation, POS tagging), multimodal feature extraction (BERT / Word2Vec / Sentence-Transformers vectorization + ResNet / EfficientNet image features), vector retrieval and similarity calculation (Milvus / Pinecone HNSW indexing + cosine / inner product measurement + Top-K screening).

[0035] Match results can be filtered by similarity threshold and displayed in a top-K format. A threshold parameter (0–1) is provided in the similarity calculation engine to automatically remove candidate vectors with similarity below the threshold. Remaining candidates are sorted by cosine similarity, with user-configurable top-K rankings (e.g., top 5, 10, or 20). In this embodiment, high-confidence labels are highlighted in the interface display, and detailed similarity scores and comparison fields for each match can be viewed with a single click.

[0036] In an optional embodiment, the system may generate a buyer / seller mismatch detail report.

[0037] After the comparison is completed, the missing items in the two sets of product lists are extracted and organized into a report according to the "Product Name / Category / Description / Image Placeholder / Price Blank / Remark Blank" fields; it supports export to Excel, CSV or JSON, and can be directly called in the custom dashboard in the BI platform (such as Power BI, Metabase).

[0038] The quotation generation module is used to fill in the quotation template according to the output of the product comparison module to generate a quotation file; the filled content includes product name, our / target product pictures, bilingual description, FOB price, MOQ and remarks. Specifically, Figure 2 As shown in the figure, the quotation generation module can automatically fill in the quotation template (Excel, PDF or API structure) based on the output results of the comparison module; insert fields such as product name, our / target product image, bilingual description, FOB price, MOQ and remarks; and beautify the template style (header / footer, logo, watermark).

[0039] In specific implementation, the quotation generation module document template engine (exceljs / docx-template dynamic field filling), brand element injection (logo, header / footer SVG or PNG embedding), format beautification and paging (automatic border, font, column width and row height adaptation). In some embodiments, an anti-counterfeiting watermark can be automatically inserted before the quotation is generated. During the document rendering stage, Office Open XML (.xlsx / .docx) or Canvas API is used to overlay a transparent QR code / random graphic on the header or invisible layer; the watermark position, transparency, size and periodic refresh are set. When the generated document is opened, the integrity of the watermark embedding can be verified to prevent illegal tampering. In some embodiments, such as Figure 4 As shown, the quotation is presented in table format, with three columns: Product Name, Our Product Information, and Target Product Information. Product descriptions can be displayed beneath each of these columns, and each description can be displayed on multiple lines. At the bottom of the table are fields for FOB price, MOQ (minimum order quantity), and notes, as well as options for switching languages ​​and exporting the quotation.

[0040] In this embodiment, after completing product comparison, the system automatically generates a detailed report of unmatched items and a standardized quotation document. The system extracts unmatched products between buyers and sellers, organizes them into reports based on key information fields, and supports exporting them in multiple formats and integrating them into the BI platform. Furthermore, the quotation generation module automatically populates quotation templates with branding elements and anti-counterfeiting watermarks based on the comparison results, generating professional quotation documents containing bilingual information, product images, and prices, thereby improving business efficiency, standardization, and anti-tampering security.

[0041] In an optional embodiment, the template engine supports Markdown and HTML output. A Markdown renderer (such as CommonMark.js) is added to the quotation generation module to replace the fields in the template and generate an .md document. The HTML+CSS rendering is exported to PDF through Headless Chrome (Puppeteer) or wkhtmltopdf to meet more flexible typesetting requirements.

[0042] The editable interface module is used to receive modification instructions for the quotation file through a visual form interface to obtain the final quotation file. Specifically, Figure 2 As shown, the editable interface module can provide a visual form-based interface on the web or mobile terminal; support real-time modification of key fields such as price, MOQ, notes, etc. in the quotation document; implement Undo / Redo, multi-version snapshot and field lock / unlock functions; and synchronize the user's editing results to the backend.

[0043] In specific implementation, the editable interface module can be used to support: Web / mobile inline table editing (React+ProseMirror), multi-version snapshots and Undo / Redo (immer manages immutable data), custom field locking, and annotation collaboration.

[0044] In this embodiment, by introducing Markdown and HTML template engines, it supports document output in multiple formats and beautiful typesetting to meet different business presentation needs; at the same time, the editable interface module provides a visual form editing interface for Web / mobile terminals, allowing users to modify quotation content (such as price, MOQ, remarks) in real time, and has version control, undo and redo, and field locking functions to ensure efficient, controllable and collaborative editing experience of quotation documents.

[0045] Optionally, quotes and edits can be made in the native mobile app. Develop mobile clients based on Flutter or React Native, invoking homologous RESTful and WebSocket services. Support for local SQLite offline caching, resumable downloads, and network recovery mechanisms are also available. Built-in push notifications provide real-time notifications of quote completion, collaborative comments, and delivery status.

[0046] The output and sending module is used to export the final quotation file into a target format file and send it to the target address. Figure 5 As shown, the output and sending module can export the final quotation file to Excel, PDF or return it through API; call SMTP or RESTful interface to send batch emails; push file links and sending status to CRM / ERP system through Webhook; and record receipts and sending logs for query.

[0047] In specific implementation, the output and sending modules can be used for multi-format export (Excel / PDF via PDFKit or Puppeteer / API JSON), SMTP / RESTful batch sending (Nodemailer + status tracking), and Webhook real-time push.

[0048] In an exemplary embodiment, as Figure 6As shown, the operation and security module is used to: deploy with Terraform / IaC and Docker / Kubernetes management systems; maintain asynchronous task queues and retry strategies through RabbitMQ / Bull; connect to Prometheus / Grafana to monitor service health and queue length, and configure alarms; enable IstioServiceMesh to implement mTLS, circuit breaking and traffic management; use JWT+RBAC and audit logs to ensure multi-tenant data isolation and compliance.

[0049] In this embodiment, through mobile development based on Flutter or React Native, combined with offline caching and push mechanisms, quotation documents can be edited at any time and collaborative reminders can be provided. The output and sending module supports multi-format export and multi-channel delivery (email, API, Webhook), and can be linked with CRM / ERP systems. At the same time, the operation and security module uses containerization and automated operation and maintenance tools, asynchronous queues, service grids, security authentication and monitoring mechanisms to ensure system high availability, data security and compliance operations in a multi-tenant environment.

[0050] Optionally, the data acquisition module includes: a website crawling submodule, a keyword crawling submodule, a file uploading submodule and a data cleaning submodule; wherein, The website crawling submodule performs headless browser deep rendering and DOM extraction based on Puppeteer or Playwright; The keyword crawling submodule is used to extract keywords based on the URLs and files input by the user; The file upload submodule is used for streaming parsing of Excel / CSV and image text recognition using TesseractOCR; The data cleaning submodule cleans, removes duplications, and standardizes the original product data, and stores it in a structured manner.

[0051] In practice, the website crawling submodule uses Puppeteer / Playwright deep rendering and multi-threaded crawling. The keyword crawling submodule supports parallel crawling, proxy pool rate limiting, and dynamic IP rotation. The file upload submodule uses Excel / CSV streaming parsing and local Tesseract / cloud-based OCR image recognition. The data cleaning submodule can remove duplicate data, standardize fields, convert units, and filter outliers.

[0052] In this embodiment, product data is efficiently acquired through website crawling, keyword crawling, and file uploading, and comprehensive information extraction is achieved by combining technologies such as headless browser rendering, concurrent crawling, and OCR recognition; the data cleaning submodule deduplicates, standardizes, and structures the original data for storage, providing an accurate and consistent high-quality data foundation for subsequent product comparison and quotation generation.

[0053] Optionally, the crawler module supports distributed crawling and dynamic proxies.

[0054] Deploy crawler microservice clusters based on Kubernetes or Docker Swarm to achieve horizontal expansion; Integrate a proxy pool (such as Scrapy-ProxyPool or self-developed), with IP rotation and User-Agent masquerading; For large-scale sites, use distributed task scheduling (Celery / Quartz) to crawl in batches and automatically retry.

[0055] In this embodiment, by deploying a distributed crawler service on Kubernetes or Docker Swarm, task parallelization and dynamic expansion are achieved; by combining proxy pools, IP rotation and User-Agent masquerading, anti-crawling mechanisms are effectively circumvented; and through distributed task scheduling and automatic retry mechanisms, the data crawling efficiency and stability of large-scale sites are improved, ensuring the high performance and reliability of the system in multi-source data collection.

[0056] Optionally, the product comparison module maps multilingual texts into vectors and stores them in a vector database through automatic translation services and BERT or Sentence-Transformers embedding models.

[0057] In this embodiment, by integrating automatic translation services and embedding models such as BERT or Sentence-Transformers, product descriptions in different languages ​​are converted into vector representations in a unified semantic space and stored in a vector database, thereby achieving high-precision cross-language product similarity calculation and comparison, and improving the adaptability and accuracy of the system in international scenarios.

[0058] Optionally, the product comparison module further includes an image feature extraction submodule based on ResNet or EfficientNet, and fuses and compares the text vector with the visual vector.

[0059] In this embodiment, by introducing an image feature extraction submodule based on ResNet or EfficientNet, the system can extract deep visual features of product images and fuse them with text vectors for comparison, realizing joint similarity analysis of images and texts, thereby improving the accuracy and robustness of the comparison results.

[0060] Optionally, the quotation generation module loads a dynamic template based on exceljs or docx-template, and automatically fills in fields, injects a logo, and beautifies the format.

[0061] In this embodiment, by loading dynamic templates based on exceljs or docx-template, the system can automatically fill in product field information, embed the company logo, and perform style beautification (such as fonts, borders, and layout), thereby efficiently generating professional, unified, and brand-recognizable quotation documents, improving quotation efficiency and customer experience.

[0062] Optionally, the editable interface module is used for online inline table editing, Undo / Redo, version snapshot and field locking functions.

[0063] In this embodiment, undo / redo, multi-version snapshot management, and key field locking functions are provided to ensure the flexibility, security, and traceability of the editing process, and improve collaboration efficiency and data accuracy.

[0064] Optionally, the output and sending module exports via Excel / PDF / API and sends in batches to a target address via SMTP or RESTful.

[0065] In this embodiment, quotation files can be exported in Excel, PDF or API formats and sent in batches to designated targets via email (SMTP) or interface (RESTful), achieving efficient and flexible file distribution and delivery.

[0066] Optionally, the operation and security module includes Terraform IaC, Docker / Kubernetes deployment, RabbitMQ / Bull task queue and Prometheus / Grafana monitoring.

[0067] In this embodiment, automated deployment is achieved through Terraform Infrastructure as Code (IaC) and Docker / Kubernetes, RabbitMQ or Bull is used to manage the task queue, and Prometheus and Grafana are combined for system monitoring and alarming to ensure the efficient operation, maintainability and stability of the system.

[0068] Optionally, the system is deployed in a cloud environment and uses MongoDB sharding or Shard Key and JWT+RBAC to achieve multi-tenant isolation and permission control.

[0069] In this embodiment, multi-tenant data isolation is achieved through MongoDB's table partitioning or Shard Key mechanism, and combined with JWT authentication and RBAC permission control, data security and access rights between different tenants are ensured as much as possible, thereby improving the security and scalability of the system.

[0070] Optionally, the system further includes an online editing module; The line editing module is configured to support Flutter or React Native mobile app breakpoint resumption and offline caching functions.

[0071] In specific implementation, online editing supports team collaboration and permission control: the online editing module integrates WebSocket (Socket.io) to achieve real-time collaborative editing by multiple people; the backend uses JWT + OAuth2 authentication, combined with a fine-grained RBAC permission model to control "who can edit / view / export"; all operations generate audit logs, supporting retrieval and rollback by dimensions such as user, IP, and time.

[0072] This implementation supports offline caching and resumable downloads for Flutter or React Native mobile apps, and enables real-time collaborative editing for multiple users via WebSocket. Combining JWT and OAuth2 authentication mechanisms with fine-grained RBAC permission control ensures the security and controllability of the collaborative process. Audit logs record all operations, enabling full process traceability and rollback, improving collaboration efficiency and system compliance.

[0073] Optionally, the operation and security module also includes integrated WAF and DDoS protection strategies.

[0074] In this embodiment, the Web Application Firewall (WAF) and DDoS protection strategy are used to defend against common Web attacks (such as SQL injection, XSS) and distributed denial of service attacks, thereby enhancing the system's network security protection capabilities and ensuring service stability and availability.

[0075] Optionally, a Webhook notification submodule is also included to automatically push the download link and status to the external system after the quotation is generated.

[0076] In practice, quotations can be pushed to third-party systems via webhooks. For example, the Output and Send modules allow configuration of one or more webhook URLs. Once a quote is generated, the system pushes the quote details, file download link, and metadata in HMAC-SHA256-signed JSON format. Retry mechanisms and receipt confirmation are supported to ensure consistent delivery to third-party systems.

[0077] In this embodiment, by configuring Webhook, quotation information can be securely pushed to the designated system, and a retry and receipt confirmation mechanism is supported to ensure that the third-party system can reliably receive and process notifications, thereby improving the automated collaboration capabilities between systems.

[0078] Optionally, an unmatched item report submodule is included for automatically generating and exporting a detailed list of unmatched products of the buyer or seller.

[0079] In this embodiment, the unmatched item report submodule automatically generates and exports a detailed list of unmatched products of the buyer or seller to facilitate subsequent supplementation, analysis and business decision-making. Optionally, it also includes: A main account and multiple sub-account management subsystems are used to share the "read" and "sent" status of overseas customers; When any sub-account performs a view or email sending operation on a target customer, the system will synchronously update the status flag in the customer views of all sub-accounts.

[0080] The system can maintain the "viewed" and "sent" status of each overseas customer based on the main account / sub-account architecture: all sub-accounts share the same customer table. When any sub-account views or sends an email, the global status of the customer is automatically updated. In the customer list or detail page, "viewed" and "sent" are prominently marked with small icons or text at the top. Status changes are broadcast through the message queue (RabbitMQ / Bull), and the front-end is refreshed in real time to avoid repeated operations.

[0081] Optionally, the quotation generation module is further configured to: Automatically add a clickable hyperlink to each product information field in the quotation document, which points to the product's detail page on the seller's or buyer's company website.

[0082] During implementation, add a clickable link to each product name or icon field in the quotation: when clicked, the front-end will jump to the product details page on the corresponding company's official website (the URL can be captured from the data acquisition module and stored in metadata); connect the web version and mobile terminal, open a new tab or embed WebView; when exporting documents (Excel / PDF), embed the link in the cell hyperlink or PDF annotation to ensure that the exported file can be directly accessed by clicking.

[0083] Optionally, the system can be connected to third-party translation services (such as DeepL, Azure Translator). For example: In the automatic translation submodule, the strategy pattern is used to encapsulate multiple translation APIs; Dynamically select the service to call based on real-time latency, cost, and translation quality (BLEU / TM match rate); Local caching of commonly used phrases and technical terms to avoid repeated calls and improve translation consistency; Divide text batches in parallel, call different APIs asynchronously, and merge the results for post-translation QA (quality assessment).

[0084] In this implementation, a policy model intelligently selects the optimal translation API (such as DeepL or Azure Translator) and dynamically schedules it based on latency, cost, and translation quality. It also supports local terminology caching to avoid duplicate calls and performs quality assessment on batches of text after asynchronous processing, thereby improving translation efficiency, consistency, and accuracy.

[0085] In an optional embodiment, the system may further include a procurement contact location module. The system can also integrate multi-source business information crawling services (Google web pages + LinkedIn API) and run according to the following process: Use a headless browser to crawl the target company's official website and industry portals to access public contact information, procurement email addresses, etc.; retrieve and pull the target company's procurement / supply chain manager information through the official Sales Navigator or third-party supplier interface; De-duplicate and standardize contact information from different sources and merge them into the user lead database; Automatically tag according to job title and salesperson tags, and perform real-time updates and permission control in the background; This module allows sales staff to directly connect with real purchasing decision makers, avoiding sending general customer service email addresses and significantly improving contact conversion rates.

[0086] This application is based on a layered modular architecture, integrating crawler crawling, multi-source adaptation, multilingual semantic matching of OCR and embedding models, an editable template engine, and batch email sending functions. It supports multiple input methods such as website addresses, file uploads, and keyword searches, and implements scenarios such as single-URL and dual-URL comparison quotations, keyword quick quotations, full-table translation, and unmatched item reports. It can also generate bilingual or multilingual editable quotation files in formats such as Excel, PDF, or API. Users can adjust fields, insert logos and headers / footers online, save and export in real time, or send in batches via SMTP / RESTful interfaces. It has the following advantages: 1. High automation: Full-link automatic crawling, OCR, translation, and comparison, significantly reducing manual intervention; 2. Cross-language accuracy: Deep learning embedding comparison supports 15+ languages ​​with an error rate of <5%; 3. Flexible editing: online / offline dual-end inline editing, Undo / Redo and version snapshots; 4. Multi-format output: one-click export to Excel, PDF, API, and batch sending; 5. High scalability: Microservices + containerized deployment, horizontal scalability and elastic scaling; 6. Multi-tenant security: MongoDB sharding + JWT + RBAC isolation strategy to ensure data and access security; 7. Easy operation and maintenance: Terraform Iac + Prometheus / Grafana full-link monitoring reduces operation and maintenance costs.

[0087] This application can be implemented in various scenarios, such as: 1. Maternal and infant industry scenario When users enter the keyword "baby sleeping bag" in the search box, the system automatically searches the top 100 sellers in the corresponding category on the German Amazon.de platform, capturing and extracting each seller's name, monthly sales volume, product information, and contact information. The results are displayed in real time in a table format and can be exported to Excel files with one click or sent to target customers via email.

[0088] 2. Pet industry scenario The user selects the French Amazon.fr platform and enters the keyword "dog leash". The system recognizes the French content and captures the structured data of relevant sellers, including the seller's homepage link, average monthly sales volume and review rating, and displays it on the interface. Users can add their favorite sellers to the follow-up list and enter the subsequent task management module.

[0089] 3. Home Furnishing Industry Scenario When a user enters the keyword "folding table," the system performs parallel crawling on multiple sites, including Amazon.com in the United States and Amazon.co.uk in the United Kingdom. It then performs vector clustering on the multilingual product descriptions on the pages, merging multiple sellers of the same product type into a single, normalized record. It then automatically identifies the product types corresponding to different languages, ultimately generating a comparison table for the user to filter.

[0090] 4. Beauty Industry Scenario When a user enters "eyeliner" on the Italian Amazon.it platform, the system will initially match the seller's information, then aggregate and filter products based on color and brand attributes, and provide users with charts analyzing the proportion of each brand, price range, and sales volume, so they can quickly identify target suppliers.

[0091] Several exemplary applications are provided below.

[0092] 1) Example 1: End-to-end Process (T-shirt Apparel Industry) The following uses the T-shirt product in the apparel industry as an example to describe the complete end-to-end process from user input to result delivery: 1. Data Input User A pastes the URL of the T-shirt product page on the official website into the web client and clicks the "Generate Quote" button. The front-end sends the URL to the business API layer via an HTTPS request.

[0093] 2. Data Capture The crawler service receives tasks from the API layer, performs deep scrolling and rendering of the target page based on Puppeteer, automatically extracts product names, specifications, images and price information, and stores the original HTML and crawled data in MongoDB.

[0094] 3. Text Preprocessing The OCR submodule recognizes SVG or Canvas text embedded in the page, merges the OCR results with HTML parsing data, and uniformly performs word segmentation, denoising, and automatic Chinese-to-English translation.

[0095] 4. Feature Vectorization The Embedding service calls the self-developed Transformer model or the OpenAI API to map multilingual text and images into 768-dimensional and 512-dimensional vector spaces respectively through ResNet50, and writes them into the Milvus vector library in parallel.

[0096] 5. Semantic comparison The comparison module calculates the cosine similarity between our product vector and the target product vector, selects the top-5 candidates with a similarity ≥ 0.8, removes duplicates through DBSCAN clustering, and re-ranks them by regional weight and score.

[0097] 6. Quote Generation The template engine automatically fills in the Excel quotation template based on the comparison results, including the product name, pictures of our and target products, bilingual product description, FOB price, MOQ and remarks, and inserts the company logo and anti-counterfeiting QR code.

[0098] 7. Online Editing Users can open the generated quotation on the web and edit key fields such as price, MOQ, and notes in real time. It supports Undo / Redo, multi-version snapshots, and field locking functions.

[0099] 8. Send in batches After the user confirms the quotation content, click "One-click Send". The system will call SMTP or RESTful API in parallel to send the quotation sheets in batches to the customer's email address, push them to the company's CRM via Webhook, and record the sending status.

[0100] 9. Feedback Logs generated throughout the entire process (such as crawler duration, comparison time, and document generation time) are reported to Prometheus and visualized in real time on the Grafana dashboard. All monitoring metrics meet the "end-to-end ≤ 5 seconds" performance target, making it suitable for large-scale concurrent B2B quotation scenarios.

[0101] This example illustrates the complete process and efficiency of this system in performing multilingual intelligent comparison and quotation for T-shirt products in the apparel industry.

[0102] like Figure 7 As shown, the process of sending quotations in batches: 1. Use Excel or PDF template to fill in the quotation content to generate the quotation; 2. Upload to object storage (S3 / OSS); 3. Obtain the file access link and generate a secure file access link through the pre-signed URL mechanism; 4. Call the sending service to send the quotation to the user via SMTP protocol in the form of an email, and simultaneously trigger the webhook to write the sending result to the CRM system in real time; 5. Record the sending status (success / failure log), and also record and track whether the user has viewed the quotation.

[0103] 2) Example 2: Keyword Quick Quote (Pet Industry) 1. User B enters the keyword "pet backpack" in the system search box. The system triggers the keyword crawler module and only performs targeted crawling on target site pages that may contain the keyword. 2. The system reuses steps 3-7 in Example 1: OCR text supplementation → multilingual embedding vectorization → semantic comparison → clustering deduplication → template engine generation of batch Excel quotations; 3. End users can view and download batch quotations for the category in an editable interface, enabling rapid coverage of pet products industry market segments.

[0104] This example illustrates how the system can achieve batch quotation capabilities for vertical categories such as pet backpacks through keyword-targeted crawling and reuse of general processes.

[0105] like Figure 8 As shown, the keyword quick quotation process: 1. The user enters the keyword, that is, enters the target keyword in the front-end search box; 2. Split the complete keyword entered by the user into semantic units (such as Chinese word segmentation and English word segmentation), and expand the synonymous expressions of the keyword; 3. Directed crawling to extract target data from filtered related websites (browser tools such as Puppeteer or Playwright can be used); 4. Feature vectorization: converting text and image data into machine-understandable vector representations and storing the vector data in the vector database MilvusSMTP / Webhook to support efficient retrieval; 5. Semantic matching and clustering: perform clustering and deduplication based on similarity thresholds and regional weights, and output the top-K most matching product pairs; 6. Generate batch quotations, fill the filtered data into the preset template (Excel / PDF), and insert brand logos (logo, header and footer); 7. Export / Download / Send: Download the quotation to your local computer or send the quotation in batches via SMTP / Webhook.

[0106] 3) Example 3: Comparing quotations for dual URLs 1. User C enters the URLs of two competitors' websites in the search box, and the system simultaneously starts two parallel crawling tasks in the crawler queue; 2. After the crawling is completed, perform OCR supplementation, vectorization, and Top-5 matching according to steps 3-5 of Example 1; 3. The aggregation module outputs the two sets of results side by side, and the template engine generates a comparison table that supports bilingual switching between Chinese and English and differentiated highlighting; 4. Users can compare and adjust prices and MOQs online and export the comparison results to Excel or PDF.

[0107] This example demonstrates the flexibility of dual-site parallel comparison, which helps users quickly and intuitively compare competitor products and quotations.

[0108] 4) Example 4: Whole Table Translation and Unmatched Item Report 1. User D uploads an English Excel price list, and the system calls a translation service (DeepL / Google API) to generate the entire Chinese list in one click. 2. The system compares the target site crawl results with the original table data to identify entries that are not matched in the site; 3. The report-generator service exports the unmatched items into a detailed report (including product name, category, image placeholder, and blank price / MOQ columns) and pushes it to the user for manual re-entry. 4. After the user completes the supplementary recording, the subsequent steps of Example 1 or 2 can be called again to generate a complete quotation.

[0109] This example demonstrates the efficiency and accuracy of the system in large-scale document translation and difference report generation scenarios.

[0110] like Figure 9 As shown, the translation process of the entire table is: 1. The user uploads a document, such as a Word or Excel file; 2. Document parsing module, extracting cell content and paragraph content in the document; 3. Call the translation service interface (you can choose DeepL API or Google Translate API, etc.) and transfer the extracted document content to the translation service interface in batches. The translation service will translate the text in the documents; 4. Translation results are merged. After the translation service returns the translated text, the system maps the translated text back to the original document structure according to the structure and order of the original text; 5. The formatting module adjusts the format of the merged translated documents, such as adjusting paragraphs and table borders, applying company logos and headers and footers, etc. 6. Translated document generation: Generate translated documents as required (e.g. generate Word / Excel / PDF); 7. Save / Download / Send: Provides users with multiple ways to obtain translated documents (e.g., download to local computer, send via email or API).

[0111] 5) Example 5: Mobile Native App Support 1. The mobile app is developed based on the Flutter framework and shares UI components and core APIs with the web app. 2. The app features a built-in offline cache module (SQLite / IndexedDB) and a resumable download mechanism, allowing users to continue editing quotes and generating documents even when disconnected from the internet or in the background. 3. Maintain data synchronization with the server through HTTPS and WebSocket, support submitting tasks while offline, and automatically upload and trigger backend processing after network recovery; 4. The mobile terminal also supports multi-language switching, online editing, export and download, and one-click access functions, achieving a seamless experience consistent with the web terminal.

[0112] like Figure 10 As shown, the web frontend uses React as the front-end framework and Tailwind as the CSS framework for style development. The mobile app is developed using Flutter, a cross-platform mobile application development framework. The web frontend can communicate via HTTPS and WebSocket, while the mobile app can communicate via the REST API. The business API layer performs user authentication / session management, interface rate limiting / log auditing, and request distribution to microservices (RabbitMQ / HTTP). Puppeteer can be used for crawling services, and the Embedding service can be used for vectorization and visual feature extraction. The data storage layer can include MongoDB (relational database), Mius (vector database), and S3 / OSS (Simple Storage Service / Object Storage Service).

[0113] This embodiment illustrates the full-function adaptability and reliability of this system on the mobile terminal, meeting the quotation needs of modern B2B salesmen anytime and anywhere.

[0114] 6) Example 6: Computer Hardware Architecture like Figure 11As shown, embodiments of the present application provide a computer hardware architecture. The system bus is the channel for data transmission and communication between various computer components. It is responsible for connecting the CPU, GPU, memory, and other external devices, enabling these components to exchange information with each other. The CPU (Central Processing Unit) is the computer's core computing and control unit, responsible for executing instructions in computer programs and performing various processing tasks such as arithmetic and logical operations. The GPU (Graphics Processing Unit) is used to handle graphics-related computing tasks, such as rendering 3D graphics and playing videos. Memory (RAM, Random Access Memory) is a component of a computer used to temporarily store running programs and data. Storage (SSD / HDD) is a hard drive, including solid-state drives (SSD) and mechanical hard drives (HDD), which is used for long-term storage of large amounts of information such as operating systems, applications, and user data. Network interfaces (such as Ethernet / Wi-Fi) are components used to connect and communicate between a computer and a network. I / O controllers (I / O, or input / output) are responsible for managing input and output operations between a computer and various external devices. USB (Universal Serial Bus) and GPIO (General Purpose Input / Output) are two common types of I / O interfaces.

[0115] The protection scope of this application is not limited to the above-mentioned embodiments. For each module, step, and functional unit in this application, a person skilled in the art can implement at least thirty equivalent or alternative implementations below without inventive work. All of the following solutions and their combinations shall be deemed as part of the protection scope of this application: 1. Single-device local deployment: All modules are integrated and run in parallel on a single computer through virtualization technology. 2. Multi-node distributed deployment: The crawler, comparison, and generation modules are deployed separately on cluster nodes.

[0116] 3. Hardware acceleration implementation: offload the embedding operation to the GPU or TPU chip for acceleration.

[0117] 4. FPGA or ASIC customization: Implement some key modules as dedicated hardware circuits.

[0118] 5. Implement core algorithms using different programming languages ​​such as Java, Python, C++, Go, or Rust.

[0119] 6. Modular containerization: Encapsulate each sub-module through containerization and use Kubernetes or Docker Swarm for orchestration and scaling.

[0120] 7. Microkernel architecture: Aggregate the "editable interface" and "quote generation" modules into a microkernel component.

[0121] 8. Use WebAssembly to move some front-end logic to the browser sandbox for execution.

[0122] 9. Trigger asynchronous tasks on demand through Serverless architecture (such as AWS Lambda and Azure Functions).

[0123] 10. Replace MongoDB with MySQL, PostgreSQL, or a cloud-native database solution.

[0124] 11. Use Redis or Memcached as a cache layer to speed up access to hot data.

[0125] 12. Replace the message queue type with Kafka, ActiveMQ, or ZeroMQ.

[0126] 13. Use different OCR engines (such as Tesseract, Baidu OCR) for text parsing.

[0127] 14. Use a different image recognition framework (such as OpenCV, TensorFlow, or PyTorch) for visual matching. 15. Use a multi-tiered local / cloud hybrid storage strategy to store documents on local NAS or cloud storage systems.

[0128] 16. Added internationalization extensions to support output in more languages ​​including Spanish, Russian, Japanese, Korean, Portuguese, etc.

[0129] 17. Replace the template engine from Excel.js to Aspose.Cells or Apache POI for document generation. 18. Generate quotation reports using Markdown or LaTeX templates and export them to HTML or PDF.

[0130] 19. Support CLI command line tool format to facilitate automated script calling.

[0131] 20. Provides two interface styles: RESTful API and GraphQL to meet different development needs.

[0132] 21. Integrate OAuth2 or SAML2.0 for single sign-on and permission management.

[0133] 22. Use Service Worker to cache static resources on the browser side to make them available offline.

[0134] 23. Replace the "keyword quotation" function with voice input and speech synthesis function.

[0135] 24. Connect with RPA robots to realize unattended one-click quotation process.

[0136] 25. Enhanced security by adding WAF, DDoS protection and security audit modules.

[0137] 26. Integrate blockchain or distributed ledger technology to ensure the immutability of quote data.

[0138] 27. Implement native App version on mobile terminals, supporting iOS / Android platforms.

[0139] 28. Introduce the micro-frontend architecture and split different modules into sub-applications for independent loading.

[0140] 29. Add big data analysis and visualization modules, and use ELK or Grafana to display key indicators.

[0141] 30. Support real-time collaboration, multiple users can edit quotations online at the same time and synchronize them in real time.

[0142] 31. Incorporate AI-assisted recommendation algorithms into the quote generation process to achieve intelligent pricing.

[0143] 32. Automatically optimize template layout and field configuration through A / B testing.

[0144] 33. Use remote disaster recovery strategies to mirror data to multi-region cloud deployments.

[0145] 1. The multilingual matching solution provided in the embodiments of this application has the advantages of high accuracy, low latency, and horizontal scalability. 2. The online editing interface solution provided by the embodiments of this application includes advantages such as ease of use, real-time preview, and multi-version management; 3. The asynchronous task queue solution provided in the embodiment of this application includes advantages such as high concurrency support, automatic retry, and current limiting alarm.

[0146] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present application can be implemented using general-purpose computer devices. They can be concentrated on a single computer device or distributed on a network composed of multiple computer devices. Alternatively, they can be implemented using program codes executable by the computer device, so that they can be stored in a storage device and executed by the computer device. In some cases, the steps shown or described can be performed in a different order than herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the embodiments of the present application are not limited to any specific combination of hardware and software.

[0147] It should be noted that the above are only preferred embodiments of the present application and do not limit the scope of patent protection of the present application. Any equivalent structure or equivalent process transformation made using the contents of the description and drawings of this application, or directly or indirectly applied in other related technical fields, are also included in the scope of patent protection of the present application.

Claims

1. A multi-language intelligent product comparison and editable quotation system, characterized by: include: Data acquisition module, product comparison module, quotation generation module, editable interface module, output and sending module; among them: A data acquisition module, configured to acquire product data of a plurality of products, wherein the product data includes product names, descriptions, and images; The product comparison module is used to: extract features from product data to obtain product vectors; calculate the similarity between product vectors; perform clustering and deduplication based on similarity thresholds and regional weights, and output the top-K most matching product pairs; A quotation generation module is configured to fill in the quotation template with the output of the product comparison module to generate a quotation document; the filled-in information includes the product name, images of our / target products, bilingual descriptions, FOB price, MOQ, and remarks; The editable interface module is used to: receive modification instructions for the quotation document through a visual form interface to obtain the final quotation document; The output and sending module is used to export the final quotation file into a target format file and send it to a target address.

2. The system according to claim 1, wherein: Also includes: The operation and security module is used to: deploy with Terraform / IaC and Docker / Kubernetes management systems; maintain asynchronous task queues and retry strategies through RabbitMQ / Bull; connect to Prometheus / Grafana to monitor service health and queue lengths, and configure alarms; enable Istio ServiceMesh to implement mTLS, circuit breaking, and traffic management; use JWT+RBAC and audit logs to ensure multi-tenant data isolation and compliance.

3. The system according to claim 1, wherein: The data acquisition module includes: a website crawling submodule, a keyword crawling submodule, a file uploading submodule and a data cleaning submodule; wherein, The website crawling submodule performs headless browser deep rendering and DOM extraction based on Puppeteer or Playwright; The keyword crawling submodule is used to extract keywords based on the URLs and files input by the user; The file upload submodule is used for streaming parsing of Excel / CSV and image text recognition using TesseractOCR; The data cleaning submodule cleans, removes duplications, and standardizes the original product data, and stores it in a structured manner.

4. The system according to claim 1, wherein: The product comparison module maps multilingual texts into vectors and stores them in a vector database through automatic translation services and BERT or Sentence-Transformers embedding models.

5. The system according to claim 4, characterized in that The product comparison module further includes an image feature extraction submodule based on ResNet or EfficientNet, and fuses and compares text vectors with visual vectors.

6. The system according to claim 1, wherein: The quotation generation module loads dynamic templates based on exceljs or docx-template, and automatically fills in fields, injects logos, and beautifies the format.

7. The system according to claim 1, wherein: The editable interface module is used for online inline table editing, Undo / Redo, version snapshot and field locking functions.

8. The system according to claim 1, wherein: The output and sending module exports via Excel / PDF / API and sends in batches to the target address via SMTP or RESTful.

9. The system according to claim 2, wherein: The operation and security modules include Terraform IaC, Docker / Kubernetes deployment, RabbitMQ / Bull task queues, and Prometheus / Grafana monitoring.

10. The system according to claim 1, wherein The system is deployed in a cloud environment and uses MongoDB sharding or Shard Key and JWT+RBAC to achieve multi-tenant isolation and permission control.

11. The system according to claim 1, wherein: The system also includes an online editing module; The line editing module is configured to support Flutter or React Native mobile app breakpoint resumption and offline caching functions.

12. The system according to claim 1, wherein: The operation and security module also includes integrated WAF and DDoS protection strategies.

13. The system according to claim 1, wherein: It also includes a Webhook notification submodule, which is used to automatically push download links and status to external systems after the quotation is generated.

14. The system according to claim 1, wherein: Also included is the Unmatched Items Report submodule, which automatically generates and exports a detailed list of unmatched products by the buyer or seller.

15. The system according to claim 1, wherein: Also includes: A main account and multiple sub-account management subsystems are used to share the "viewed" and "sent" status of overseas customers; When any sub-account performs a view or email sending operation on a target customer, the system will synchronously update the status flag in the customer views of all sub-accounts.

16. The system according to claim 1, wherein: The quotation generation module is further used for: Automatically add a clickable hyperlink to each product information field in the quotation document, which points to the product's detail page on the seller's or buyer's company website.

Images