The Text Cropping System is a machine learning and computer vision program that automatically extracts individual entries from a historic dictionary and saves each one as its own clean, labeled image. Sponsored by Dr. Amy Giroux, the project tackles the digitization of Samuel Johnson’s Dictionary of the English Language—one of the most influential works in English literary history, and one whose more than 42,000 entries had previously been cropped by hand, an enormously time-consuming task. The software learns to recognize the visual formatting cues that mark where one entry ends and the next begins, then crops each headword along with its definition, straightens any skewed text, and strips away extraneous material like page numbers and guide words. Each cropped entry is saved as an image named for its headword, preserving the look and feel of the original printed page while making the dictionary far easier to store, search, and share. Beyond this single dictionary, the team designed the tool as a low-cost, scalable solution applicable to digitizing all kinds of historical documents at scale.
Senior Design Project — Samuel Johnson’s Dictionary Online