OLSYS Ltd provides full-service solutions for mid-market and enterprise organizations.
As an enterprise software development company, we are building long term partnerships helping our clients accelerate their digital experiences with reasonable IT investments.
Our tailored approach, e-commerce focus, and flexible solutions allow us to design, develop, and deliver scalable, integrated commerce platforms that drive profits and boost the business.
15+ years of experience, 100+ projects, 50+ specialists
About the Client
Our client is a leader in Process Discovery and Mining Solutions, helping businesses transform the way they monitor, analyze, and optimize their processes. The company offers precise tools for continuous control and improvement, delivering significant time and operational cost savings to its clients.
About the Project
This is a unique opportunity to work on product development within a service company. The product provides recommendations for improving personal work environments by analyzing user behavior and offering actionable insights. The project involves extensive work with text analysis and image processing.
The primary challenge is detecting various entities in OCR-processed screenshot texts (e.g., names, addresses). Named Entity Recognition (NER) in screenshots differs from NER in raw text, as screenshots typically contain fragmented text that is not interconnected and cannot be read sequentially (left to right, top to bottom). Instead, 2D coordinate information is used, as positional encodings are insufficient. In addition to textual and positional data, the visual features of the screenshot — such as layout and font — are considered, providing a third input source for the model.
Domain: Screenshot/document layout understanding, Named Entity Recognition (NER).
English level: Upper-Intermediate