Optical character recognition, often shortened to OCR, identifies characters or text sequences in visual input such as scans and photographs. The output can be used for search, editing or later document processing. Modern systems may combine visual recognition with language-based predictions.
Recognising plausible text is different from faithfully recovering every character. Image quality, layout, unusual names and missing details can cause errors. Language predictions may produce convincing words that were not actually present, so important outputs need appropriate verification.
ELI5
OCR helps a computer turn a picture of writing into text it can work with. It tries to read the marks in the image.
For example, a scanned receipt could be turned into searchable text. If a number is blurred, the system may read it incorrectly even when the rest looks right.
Does correct-looking OCR prove that every word was recovered?
No. A system can produce plausible text despite missing or misreading visual details.
Is recognising text the same as understanding a document?
No. Converting visible writing to text is one step; interpreting its meaning and retaining it accurately are separate tasks.
