Case Study

Column Digitizes 50 Thousand Text Documents per Year With Computer Vision

Locating and Transcribing Text from Images for Public Notice Digitization

Column is the first collaborative public notice platform helping publishers, governments and legal services work together to inform their communities. Column digitizes the process of managing public notices and helps ensure compliance with public notices by transforming images of text (newspapers, PDFs, documents) into usable digital data. Column saw an opportunity to build a system for replacing the manual steps of finding and transcribing public notice information by creating machine readable information from newspapers.

45,000+ Monthly API Calls Digitizing Text-Based Public Notices Into Usable Data

In less than 2 weeks, using Roboflow’s self-serve annotation and labeling product, Roboflow Annotate, Column labeled 3,000+ images with the help of external freelance labelers managed in Roboflow. Utilizing roles, permissions, job assignments, notifications, and instructions to scale labeling outside of their organization.

Using Roboflow Train to create a hosted model in one-click and having results quickly thanks to high-performance compute from AWS, Column called the Roboflow API to detect public notices within documents and used an OCR endpoint to translate the text automatically. Roboflow’s autoscaling inference API includes load balancing, supports burst and no burst, and is always on with a simple pay-for-use model which requires no custom engineering to configure.

Manual hours saved using computer vision per year
Increase in speed of transcription
Documents per year read with computer vision