Newspaper Ads Engine
Our AI software for turning Pakistani newspaper advertisements into searchable, database-ready information for websites and services.
We use a trained YOLO model to detect advertisement sections on newspaper pages. More than 60,000 ads are awaiting structured-content generation. Our current workflow uses EasyOCR, PaddleOCR and Gemma, supported by a dedicated development PC with a 16 GB GPU. We are developing the structured-data layer so detected ads can become useful records, with review against the original advertisement.
- Designed for categories, titles, descriptions and relevant details extracted from each advertisement
- Next step: evaluate larger AI models for extraction accuracy, consistent structure and cost per ad
- Intended for job portals, tender boards, classifieds platforms and other subscribed clients