|
Vol.14, No.3, August 2025. ISSN: 2217-8309 eISSN: 2217-8333
TEM Journal
TECHNOLOGY, EDUCATION, MANAGEMENT, INFORMATICS Association for Information Communication Technology Education and Science |
Training a Model for Automated Information Retrieval from the Internet
Stanislav Dakov, Megi Dakova
© 2025 Stanislav Dakov, published by UIKTEN. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. (CC BY-NC-ND 4.0)
Citation Information: TEM Journal. Volume 14, Issue 3, Pages 2588-2598, ISSN 2217-8309, DOI: 10.18421/TEM143-61, August 2025.
Received: 23 November 2024. Revised: 27 March 2025.
Abstract:
This paper presents an end-to-end visual pipeline for automated extraction of structured product data—specifically product titles, images, and prices—from e-commerce websites. The approach leverages the You Only Look Once (YOLO) v8 object detection model to identify and localize key product elements directly from rendered webpage screenshots. A hybrid dataset generation methodology was developed, combining initial manual annotation with a scalable Java-based crawler that identifies Document Object Model (DOM) elements using Cascading Style Sheets (CSS) selectors and calculates normalized bounding boxes for YOLO-compatible training. A dataset of over 30,000 annotated screenshots from 30 unique websites was compiled to train the detection model. Evaluation on 10 previously unseen websites demonstrated robust generalization: 100% detection accuracy for product images and titles, and 60% for prices, resulting in an overall success rate of 86.7%. These results highlight the feasibility of vision-based extraction methods as a scalable alternative to traditional DOM-based scraping, particularly in contexts with diverse or dynamically generated web layouts. The paper concludes with recommendations for improving price detection accuracy through further model fine-tuning and multimodal learning techniques.
Keywords – Model, information, internet, automated retrieval. |
|
----------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------- |