tiny text ahead move up quality ocr
play

TINY TEXT AHEAD! Move up! Quality OCR A TANGO OF AVAILABLE - PowerPoint PPT Presentation

TINY TEXT AHEAD! Move up! Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo, Digital Lifecycle Lead Mira Basara, Ingest Collection Specialist Cornell University Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo,


  1. TINY TEXT AHEAD! Move up!

  2. Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo, Digital Lifecycle Lead Mira Basara, Ingest Collection Specialist Cornell University

  3. Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo, Digital Lifecycle Lead Mira Basara, Ingest Collection Specialist Cornell University

  4. Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo, Digital Lifecycle Lead Mira Basara, Ingest Collection Specialist Cornell University

  5. Who among us… …are OCR nerds?

  6. History ‐ OCR at Cornell D‐Lib Magazine , October 1996, http://mirror.dlib.org/dlib/october96/cornell/10chapman.html

  7. evolutions… TextBridge

  8. our current practice CHALLENGES AND STRATEGIES

  9. TextBridge 2007 ABBYY FR 2019

  10. TextBridge 2005 ABBYY FR 2019

  11. Latin Latin, Greek

  12. Inverted, cropped, deskewed image Original image

  13. OCR from Original OCR from inverted, cropped

  14. Character fill – pattern training helps

  15. Fraktur OCR of Fraktur (as found in older German texts) is very difficult. • For many characters, there are rules that govern the form of a character depending in its position. • Other characters that are interpreted differently may resemble each other closely.

  16. Fraktur OCR ‐ Google Cloud Vision

  17. opportunities DECISION POINTS

  18. Points of opportunity Re‐OCR

  19. Forced decision point Upgraded

  20. Testing alternatives

  21. Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo, Digital Lifecycle Lead Mira Basara, Ingest Collection Specialist Cornell University

  22. Quality OCR A TANGO OF AVAILABLE RESOURCES Michelle Paolillo, Digital Lifecycle Lead Mira Basara, Ingest Collection Specialist Cornell University

  23. Discussion navigate to www.menti.com And use code 97 04 40

Recommend


More recommend