Optical Character Recognition
Computer recognition of visual text. Subclass of pattern recognition. Linked to handwriting recognition.
Entity
Q167555 — Wikidata identifier for optical character recognition
Domain
Software feature • Document digitization • Algorithm
Parent Class
Pattern recognition (subclass relationship)
Sibling Systems
Natural language processing • Optical music recognition • Handwriting recognition
Postal Application
USPS Delivery Point Barcode Reader (DPBR) — reads ZIP+4 codes at 30,000 pieces per hour
Resolution Threshold
Minimum 200 DPI for handwritten address recognition; 300 DPI for degraded ink
Error Rate Target
< 0.1% misclassification at first pass; human review triggers at 99.9% confidence
Generation Alpha (1960s–1970s)
Early experimental OCR systems developed by MIT Lincoln Laboratory. First successful reading of typewritten addresses using template matching algorithms.
Source: Postal History Branch Q1516929
Generation Beta (1980s)
USPS Automation Data Capture System (ADCS) deployed nationwide. Neural network classifiers replace rigid templates. Dwell time reduced from 12 to 3 seconds per piece.
My first shift: Boston District Facility, 1987
Generation Gamma (1990s–2000s)
Delivery Point Barcode Sorter integrates OCR with laser marking. Every envelope receives a unique 10-digit code. Throughput exceeds 50,000 pieces per hour.
Subclass: Text Digitization (P31)
Current State (2020s)
Deep learning models trained on 40 years of degraded handwriting. Real-time correction loops between vision system and mechanical actuators. Monthly views on OCR research: 4,102 and climbing.
Velocity Score: 1.0 (rising trend)