Remove punctuation and characters

Choose a narrow removal rule before cleaning

Start by naming exactly what should disappear. If you only need to remove number signs from a copied label list, choose Selected characters and enter the number sign. If you need to remove separators that are all dashes, choose Dashes. A broad punctuation rule affects more than the visible marks in your first line, so use it only when every affected category makes sense for the task.

The selected-character field is a list of literal characters, not a search pattern. Typing two marks means “remove either of these characters,” not “remove this exact two-character phrase.” Use Find and replace with an empty replacement if the full phrase or sequence must match as a unit. This distinction matters for markers such as [DRAFT], where deleting the bracket characters everywhere is different from deleting a specific tag.

Test on a small block containing an expected removal and a mark that must stay. Compare the results before applying a rule to the entire input. A count of removed characters is useful only when the rule is correct. In a data export, a single deleted decimal mark, delimiter or sign can change a value, so keep the source unchanged until the result has been reviewed.

Keep meaningful hyphens and apostrophes when removing punctuation

Try this line with Punctuation selected, keeping hyphens and apostrophes on. The comma, colon, exclamation mark and final period can go while the internal marks in the words stay.

Ready, set: go! Don’t drop the two-part code.

Review the text after changing each exception. A hyphen can be part of a compound, a negative number or an identifier. An apostrophe can be part of a contraction or a name. Preserving those marks generally produces more readable prose, but it is not a universal rule. Decide what the receiving system wants and inspect the actual values.

The punctuation setting is based on Unicode punctuation, not the visual idea of “any non-letter.” A currency symbol, emoji or other symbol may be outside that category. If one particular character remains, use the specific mode or literal selection rather than assuming the converter missed all punctuation. If you cannot identify the character, inspect the original with the invisible character inspector or compare a pair by character.

Decide between deletion and replacement with a space

Removing the dash from north—south produces northsouth if the mark is simply deleted. If the dash acts as a separator, enable Replace with a space to preserve a boundary. The same question arises with a slash, a colon or a comma between short labels. A removal that yields one continuous word can look neat but change how the result is searched or read.

Replacement can also create repeated spaces when the mark already has spaces around it. Review the result before deciding whether to collapse those gaps. Use Clean whitespace on a working copy when extra spaces are unwanted, and leave indentation or fixed-width columns alone. The character remover cannot tell whether an apparent gap is an important field separator.

Do not use blanket punctuation removal to split data into columns. If commas or tabs carry structure, read the table with the CSV and TSV converter or the target importer's parser first. A delimiter in a record and the same character inside a quoted value have different roles. Select the text field that actually needs cleaning rather than applying one rule across the entire export.

Understand what the dedicated digit and dash modes can affect

Decimal digits include digit characters from more than one writing system. This is useful when you deliberately want to remove numerical digits from pasted text, but risky when a name, part number or date must keep them. A superscript or a fraction may not be a decimal digit in that category. If you want to remove one exact visible symbol, use Selected characters and verify the code point in a small sample.

Dashes also have different meanings. A short hyphen in a name, a long dash between clauses and a minus-like mark in a numeric expression can all be significant. Test the dedicated dash mode on the kinds of marks present in your real source before copying the result. If you want a long dash to become a simple hyphen rather than disappear, try the punctuation options in Normalize Unicode instead.

Use data validation at the destination if the result will be interpreted as an identifier or numeric value. Removing punctuation from an amount can turn two different values into the same text. Removing digits from a reference can destroy the only unique part of it. The tool follows the selected rule; it does not verify that the cleaned result is a valid number, filename or key.

Clean visible labels separately from machine values

A report may contain a column for a display label and a column for the actual code. The label can tolerate typographic cleanup that would break the code. Work with the label values only, preserve the original keys, and check a pair of entries whose labels differ by punctuation. If removing marks makes two different labels look identical, decide whether that collision is acceptable before publishing.

URLs and email addresses are particularly sensitive to this distinction. A period, slash, at sign or hyphen can be essential. If the real job is to collect visible addresses, use Extract links and emails and inspect the results instead. If the job is to turn an existing title into a separate machine-friendly label, use the slug converter and keep the original title.

If you are preparing a public example, replace sensitive values deliberately rather than assuming removal of digits or punctuation anonymizes the text. Names and recognizable fragments can remain. Use a fictional example with the same kinds of characters so another person can reproduce the behavior without receiving the original content.

Review a before-and-after pair before accepting the output

Keep the original text. Try one expected match, one mark that should remain and one situation where deleting a separator would merge two words. Check the result in the actual destination, especially if the destination also interprets punctuation or trims whitespace. If it rejects the cleaned value, find the exact difference instead of widening the removal rule blindly.

Use Compare two texts in character mode for a short passage. Inspect each deletion or replacement and check that the surrounding text did not lose meaning. For a longer document, use line mode to identify affected sections and review a representative selection from each. The guide to reviewing text edits explains how to retain the original and document what you intended to change.

Write the cleanup rule clearly when passing the result on: for example, “Removed the selected hash character, preserved other punctuation, replaced removals with spaces.” A reviewer can repeat that operation and inspect exceptions. Keep the processed text separate from the source until its spelling, field boundaries and readability have been checked.

Inspect an invisible difference before trying to delete it

Some apparently missing separators are invisible Unicode marks rather than visible punctuation. Do not guess which character you are looking at from the gap alone. Find the smallest affected pair, use the inspector to identify a supported mark, and choose its targeted cleanup only if that mark is unwanted. A joiner or directional mark can be meaningful to a language or emoji sequence, so a result with fewer invisible characters is not automatically a better result.

Check a deletion that changes the meaning of a value

A period in a sentence and a period inside a version number are the same character but serve different purposes. If you only need to remove final punctuation from a set of labels, a broad character rule can also change values such as v2.1. Work on a copy of the label column, and inspect a version number or amount before deciding that a punctuation-only cleanup is safe.

Likewise, if a dash separates two labels, replace it with a space only when the remaining words should be read separately. If the receiving system expects a particular delimiter, use a format-aware conversion instead of treating every occurrence of a character as disposable.