After recognition, the text may be edited like typed text rather than locked as a final result

What it is

Once ink is recognised as characters and committed, the result should enter the ordinary text channel: caret, selection, keyboard, and speech can all change it. It must not become an uneditable picture or a one-shot recognition snapshot. That is post-editing. Swapping a candidate beside the ink during writing is in-place correction inside the recognition loop, before commit.

Why it happens

Commit turns recognizer output into document characters, with font, encoding, and search. If commit rasterises and throws the characters away, post-editing is another stroke or a full re-recognition—the cost jumps to redo. Editable means the document model treats the block as a text run: backspace deletes a character, a tap places a caret, a selection can be replaced from the keyboard. A handwriting origin should not make second-class text—the same undo, spell-check, and copy should apply. Locking is often excused as “keeping the handwritten look”: look can stay as an ink layer, but the character layer must stay live. Read-only recognition shows up in scan archives, or when “convert” is implemented as export rather than an in-document state. How the caret aligns on a former ink location depends on ink–glyph correspondence; without it, editing is still possible as a whole block, just not by tapping a particular glyph.

Studying it

Tasks: after recognition, change one character, insert in the middle, select three characters and replace from the keyboard. Compare editable text, a locked bitmap, and whole-block re-recognition only.

Independent variables: post-commit representation (text run / bitmap / mixed), whether the caret can land inside the block. Dependent measures: time to change one place, abandon-and-rewrite, whether copy to the outside is real text.

Separate “looks like type” from “is characters”.

Where it stops holding

Signatures and markup that must keep ink for legal effect may omit a character layer, or use it only as an index; “do not lock” then does not apply to the look layer. A read-only document (a handout someone sent) cannot be edited anyway. Putting the result only on the clipboard while the document stays ink is not in-document post-editing.

Applying it

• After convert, default to text that can take a caret. The ink layer may be hidden or stacked; it must not replace characters.
• Keyboard, speech, and spell-check treat this run like any other.
• Convert is an in-document state, not a one-shot export.
• How to check: after recognition, tap mid-line; typing should insert, not fail as a whole-block replace. Copy into a plain-text editor should be characters. Hide the ink layer; the words remain. Against a locked bitmap, post-edit should be clearly shorter.

Related

• Same group: C5.15.2 Keep ink–text correspondence to revert to the original hand · C5.15.3 Batch correction after writing costs more than per-character confirm · C5.15.4 Reflow after convert can wreck spatial layout
• Adjacent: C5.05 Handwriting Recognition and Correction · C5.16 Switching Between Handwriting and Typing
• Search: post-recognition editing · committed text · handwriting to text