People searching translation punctuation QA, double spaces translation, non-breaking spaces CAT tool, quotation marks localization, translation typography check, or spacing errors in translated documents are looking for a mechanical cleanup pass that sits after semantic translation but before delivery. These defects are small enough to escape ordinary reading and visible enough to make a finished translation look careless or break line behavior.
A fast translation workflow therefore runs a focused whitespace, punctuation and symbol pass. Modern CAT and QA systems can flag punctuation differences, whitespace, capitalization, hidden characters, and formatting issues. Editors can also show non-printing characters or offer explicit non-breaking spaces, non-breaking hyphens, en dashes and em dashes. The human task is to distinguish target-language typography from accidental source copying.
This article owns one narrow reader job: clean target-side spacing, punctuation and typographic symbols without reopening settled translation meaning. It is different from target-only fluency review and different from QA-profile setup. Here the semantic work is mostly done. The pass isolates surface mechanics so they can be corrected quickly and consistently.
Quick answer
After meaning is stable:
- enable hidden characters if useful;
- filter/search likely mechanical errors;
- check spaces around punctuation;
- remove doubles/trailing spaces;
- verify quotation marks/apostrophes;
- verify dashes/hyphens/minus signs;
- check non-breaking spaces/hyphens where required;
- confirm target-language punctuation style;
- rerun QA.
Do not force punctuation to match source one-for-one.
Target-language typography has its own rules.
Why mechanical cleanup deserves its own pass
During translation, attention is occupied by:
- meaning;
- terminology;
- syntax;
- context.
A double space is low priority.
A wrong quote mark is low priority.
Trying to monitor every micro-feature while translating can slow the high-value linguistic decisions.
A later mechanical pass batches them.
Mode specialization improves speed.
Whitespace is invisible until it is not
Spaces can affect:
- line wrapping;
- search;
- rendering;
- file comparison;
- QA;
- layout.
Common problems:
- double spaces;
- missing spaces;
- trailing spaces;
- leading spaces;
- non-breaking spaces replaced;
- accidental tabs.
Show hidden characters when diagnosing.
Worked example 1: double space
Target:
Click Save to continue.
Easy to miss in proportional font.
QA catches it.
Fix mechanically.
Do not reread entire sentence semantically.
Worked example 2: missing space near tag
Target export becomes:
ClickSave
because a formatting tag consumed visible separation.
Hidden-character view reveals no space.
This is structural typography.
Leading/trailing spaces
Some file formats need them.
Others do not.
Do not remove blindly.
If CAT tool automatically manages leading/trailing whitespace, understand that setting.
A space may separate concatenated strings.
Worked example 3: concatenated software strings
String A:
Hello
String B:
world
Runtime concatenates with source-space logic.
If translator trims a required leading space in B, output becomes:
Helloworld.
This is why source context matters even in a mechanical pass.
Non-breaking spaces
Non-breaking space prevents line break between items.
Common uses depending on language/style:
- number + unit;
- initials + surname;
- short preposition + following word;
- currency;
- dates.
Use target-language rules.
Do not insert because source has one if target convention differs.
Worked example 4: number and unit
Target:
20 kg
Style requires no line break between 20 and kg.
Use non-breaking space.
This is typography, not semantic translation.
Non-breaking hyphens
Useful when hyphenated identifier should not split across line.
Examples:
- product code;
- compound name;
- phone prefix.
Use only where style/layout requires.
Ordinary hyphen versus en dash
Common distinction:
- hyphen joins words;
- en dash marks ranges/relationships in many styles.
Source:
10-15
Target style may prefer:
10–15
Do not treat every hyphen as interchangeable.
En dash versus minus sign
Range dash and mathematical minus are different symbols.
-5 °C
should ideally use correct minus sign if technical style requires.
Do not turn negative value into range punctuation.
Em dash
Languages differ in:
- em dash;
- en dash;
- spacing around dash.
Follow target style.
Do not mechanically copy source dash spacing.
Quotation marks
Quotation conventions differ widely:
- “…”;
- ‘…’;
- « … »;
- „…“.
A translation can preserve words but use wrong target quotation style.
This is a classic target-side cleanup issue.
Worked example 5: nested quotation
Source style:
“He said ‘stop’.”
Target style may reverse quote hierarchy.
Check both outer and inner marks.
Automation can help, but context matters.
Apostrophes versus quotation marks
Straight ' may be used as:
- apostrophe;
- single quote;
- prime.
Curly apostrophe may be preferred.
Do not globally replace without checking role.
Worked example 6: foot/inch marks
Source:
5′ 6″
These may represent feet/inches.
Replacing with curly quotes can be wrong.
Use prime symbols where style requires.
This is why symbol cleanup needs context.
Ellipsis
Three periods ... versus ellipsis character ….
Project style determines.
Do not mix.
Check spacing around ellipsis.
Colon and semicolon spacing
French and other languages can differ from English spacing conventions.
Target punctuation is language-specific.
QA that compares source punctuation literally can create false positives.
Use target rules.
Question/exclamation marks
Some languages use:
- inverted opening punctuation;
- different spacing.
Ensure target complete.
Do not preserve source punctuation shape if language requires more.
Full stops in UI labels
Source button may include period.
Target product style may omit.
This is not necessarily a mismatch error.
Use UI style.
Punctuation in headings
Style card can specify:
- no final punctuation;
- question headings;
- sentence case.
Mechanical pass catches inconsistent endings.
Punctuation in bullets
Choose consistent list style:
- full sentences with periods;
- fragments without periods.
Do not mix unless source function requires.
Worked example 7: list inconsistency
- Install the driver.
- Restart computer
- Verification
Target-only/list pass may catch grammar.
Punctuation pass catches inconsistent period.
Coordinate with style.
Commas and decimal separators
Comma can be:
- punctuation;
- decimal separator.
Do not run blind comma replacement.
Numeric values need separate rules.
Thousand separators and spaces
Non-breaking spaces may represent thousands grouping.
Do not collapse:
1 500
to:
1500
without style decision.
Parentheses and brackets
Check paired structure:
- ();
- [];
- {}.
Translation can lose one bracket.
QA can detect.
Brackets may surround tags/placeholders.
Do not fix mechanically if structural.
Worked example 8: missing closing parenthesis
Target:
(see Section 4.2
Easy.
Fix.
But if source reference structure changed, verify cross-reference separately.
Slashes
Slash can mean:
- alternative;
- per;
- path;
- date.
Spacing conventions vary.
Do not normalize every slash identically.
Ampersand
Source may use & in product/UI.
Target prose may prefer word.
Brand names may require &.
Context controls.
Degree sign
20°C versus 20 °C.
Style determines spacing.
Do not confuse with superscript zero.
Multiplication sign
5 x 10 versus 5 × 10.
Technical typography may require ×.
Do not use letter x if math meaning matters.
Micro symbol
µg versus ug.
Some systems cannot display µ.
Medical content may require exact unit.
Project encoding and style matter.
Trademark symbols
® ™ ©
Usually protected.
Do not add or remove without brand/legal instruction.
Symbol pass can verify consistency.
Currency symbols
Symbol position and spacing are locale-specific.
But currency identity must remain correct.
Mechanical typography cannot decide currency conversion.
Hidden characters
CAT editors may show:
- space dots;
- tabs;
- line breaks;
- non-breaking spaces.
Enable selectively.
They can clutter visual reading.
Use during mechanical pass.
Carriage returns/line breaks
Manual line break can be intentional:
- address;
- poem;
- UI;
- label.
Do not remove as whitespace noise.
Differentiate paragraph structure from accidental breaks.
Tabs
Tables exported from plain text may rely on tabs.
Do not convert all tabs to spaces unless safe.
Repeated punctuation
Errors:
..;,,;!!.
Some are intentional:
- ellipsis;
- expressive dialogue.
QA flags candidate.
Human decides.
Source punctuation versus target punctuation
A punctuation mismatch warning asks:
different.
It does not say:
wrong.
Target language may require difference.
Mechanical pass should enforce target convention.
Worked example 9: colon
Source heading:
Note:
Target style:
Note
No colon.
Keep target style.
Ignore/resolve warning if legitimate.
Worked example 10: French quote spacing
Target uses guillemets with non-breaking spaces.
A source-mirroring reviewer may “fix” them incorrectly.
Locale style wins.
Build a typography style mini-card
For recurring project:
- quotation marks;
- dash style;
- ellipsis;
- number-unit spacing;
- list punctuation;
- heading punctuation.
This reduces repeated decisions.
QA categories
Run:
- whitespace;
- punctuation;
- capitalization;
- repeated word;
- forbidden characters.
Then review only flagged segments.
Use regex for common client rules
Examples:
- double spaces;
- straight quotes;
- space before colon;
- prohibited hyphen.
Test carefully.
Regex can create false positives.
Search before replace
Find all candidate punctuation.
Inspect.
Then batch-correct safe patterns.
Do not Replace All blindly.
Worked example 11: straight apostrophe in code
Global curly apostrophe replacement breaks code/string.
Scope search to human-readable target.
This is why structural context matters.
Mechanical pass after copy/paste
Copying from external references can introduce:
- nonstandard spaces;
- smart quotes;
- hidden formatting.
Run cleanup.
Mechanical pass after reviewer Word edits
External reviewer may introduce:
- different quotation style;
- double spaces;
- non-breaking spaces lost.
After reimport, run mechanical QA.
Mechanical pass after MT
MT often produces reasonable punctuation but may mirror source style.
Do not assume locale typography correct.
Mechanical pass after term replacement
Batch replacement can create:
- double spaces;
- missing spaces;
- duplicated punctuation.
Run pass afterward.
Mechanical pass after source update
X-translated/reused content can mix old/new punctuation styles.
Check seams.
Capitalization
Though not punctuation, surface pass often includes:
- sentence initial;
- heading case;
- brand case.
Use style card.
Do not lowercase acronyms.
Worked example 12: title case imported to sentence-case locale
Source headings title case.
Target style sentence case.
Normalize.
This is a target typography rule.
Sentence spaces
Some legacy documents use two spaces after period.
Target house style may use one.
Use controlled replacement if file is plain prose.
Avoid code/monospaced contexts.
Non-breaking line behavior
Preview representative:
- number/unit;
- personal name;
- short label.
Ensure non-breaking characters function.
Typography and accessibility
Fancy symbols should remain supported by:
- font;
- screen reader;
- platform.
Do not replace words with obscure symbols merely to save space.
Typography and fonts
Correct symbol may render as missing glyph.
Preview final font.
Use supported characters.
Typography and encoding
Legacy systems may not support certain Unicode punctuation.
Project file requirements override ideal typography.
Use safe target symbols.
Typography and RTL
Punctuation placement can be visually confusing.
Use bidi-aware editor and preview.
Do not manually reorder logical punctuation based only on screen appearance.
Typography and CJK
Full-width/half-width punctuation may matter.
Use locale and platform conventions.
Do not apply Western spacing rules.
A dedicated 10-minute cleanup pass
For medium document:
- run QA;
- fix double spaces;
- review punctuation mismatches;
- inspect quote style;
- inspect dashes/hyphens;
- check non-breaking spaces;
- final spellcheck.
This can improve finish dramatically.
The “do not reread meaning” rule
If mechanical edit could change meaning:
- minus sign;
- decimal separator;
- legal punctuation;
stop and verify.
Otherwise keep pass mechanical.
Mode separation creates speed.
Punctuation can change meaning
Examples:
Let’s eat, Grandma. Let’s eat Grandma.
This famous pattern illustrates why punctuation is not always cosmetic.
Do not edit blindly.
Hyphen can change meaning
Example:
re-sign resign
Mechanical symbol carries lexical meaning.
Verify.
Dash can mark negative
A hyphen-like glyph may be minus sign.
Do not “beautify” without context.
Apostrophe can mark possession/contraction
Removing can change word.
Spellcheck may help.
Quote marks can distinguish direct speech
Wrong nesting can confuse who said what.
This is more than typography.
Mechanical severity
Critical
- minus/decimal;
- bracket/tag structure;
- punctuation changing legal/safety meaning.
Major
- wrong quotation nesting;
- missing line break;
- incorrect non-breaking structure.
Minor
- double space;
- cosmetic dash style.
Prioritize.
Typography review in final PDF
Some errors appear only after rendering:
- missing glyph;
- awkward line break;
- stranded punctuation.
Preview final output.
Typography review in web UI
Check:
- wrap;
- truncation;
- icon adjacency.
Surface mechanics interact with layout.
Typography review in subtitles
Check:
- italics tags;
- line breaks;
- punctuation;
- speaker dash.
Timed-text style differs from prose.
Typography review in tables
Check:
- alignment;
- unit spacing;
- decimal separators;
- header punctuation.
Dense data amplifies visual inconsistencies.
Typography review in legal documents
Follow client/legal style.
Do not modernize punctuation casually if formatting conventions have legal significance.
Build a clean-target export
Sometimes easiest is target-only DOCX/PDF.
Run punctuation tools there.
Then ensure corrections return to CAT/TM if workflow requires.
Do not leave final changes outside reusable assets.
The final mechanical release gate
Before delivery:
- no unexplained QA punctuation errors;
- no accidental double spaces;
- quote style consistent;
- dashes/hyphens correct;
- non-breaking spaces used where required;
- symbols render correctly;
- hidden characters intentional.
This is a small pass with visible quality payoff.
The deeper principle: separate typographic correctness from linguistic correctness without pretending they are unrelated
Translation meaning comes first.
Presentation still affects comprehension and professionalism.
A focused surface pass lets the translator fix predictable mechanical defects quickly because the semantic decisions are already stable.
The rule is:
review the target as text mechanics now, and reopen meaning only when a symbol actually carries meaning.
Advanced practice: build a target typography profile
For recurring work, record a compact typography profile:
- outer/inner quotation marks;
- apostrophe style;
- dash style;
- range separator;
- ellipsis form;
- number–unit spacing;
- colon/semicolon spacing;
- list punctuation;
- non-breaking space rules.
This profile can feed QA and final proofreading.
It prevents every translator from deciding typography independently.
Worked example 13: range versus subtraction
Source table:
5-3
Is it:
- range 5–3?
- subtraction 5−3?
- code 5-3?
Do not normalize symbol until function known.
Mechanical pass still requires classification.
Worked example 14: date separator
Source:
2026-09-19
This may be ISO date.
Do not replace hyphens with en dashes.
Symbols can be structured data.
Worked example 15: phone number hyphen
Source:
555-0100
Hyphen belongs to formatting.
Do not convert to en dash as range punctuation.
Worked example 16: ISBN
978-1-2345-6789-0
Hyphens are identity separators.
Do not typography-normalize.
Worked example 17: markdown
-- or * may be syntax.
A punctuation replacement can break markup.
Scope mechanical cleanup to rendered prose.
Worked example 18: command line
--help
Two hyphens are functional.
Do not convert to em dash.
Structured text needs exclusion.
Worked example 19: negative sign
Source:
-0.5
A “smart dash” tool turns hyphen into en dash.
Wrong.
Use proper minus if style requires.
Worked example 20: quotation inside code
"status":"ok"
Curly quotes break JSON.
Do not smart-quote code.
This is a classic automation failure.
Create safe zones and no-touch zones
Safe mechanical cleanup:
- ordinary prose.
No-touch/inspect carefully:
- code;
- URLs;
- IDs;
- formulas;
- paths;
- markup;
- data tables.
Scope reduces risk.
Regex boundary rules
A double-space regex is usually safe in prose.
A hyphen replacement regex is rarely globally safe.
Classify rule by risk.
Mechanical QA hierarchy
Tier 1: highly safe automation
- double spaces;
- trailing spaces in ordinary prose.
Tier 2: review candidates
- straight quotes;
- repeated punctuation;
- space before punctuation.
Tier 3: context-sensitive
- hyphens/dashes;
- minus signs;
- slashes;
- apostrophes.
Do not automate Tier 3 broadly.
Hidden spaces around tags
Tags can make:
word<tag> word
look correct while exported string has:
wordword.
Check source/target boundaries.
Use preview/export sample.
Non-breaking space misuse
Too many non-breaking spaces can prevent natural wrapping.
Do not replace every ordinary space in short phrases.
Use where line break would be undesirable.
Language-specific NBSP examples
Some languages use non-breaking spaces before certain punctuation.
Others do not.
Project style and locale matter.
Do not copy another locale’s typography profile.
Narrow no-break space
Some typography uses narrow no-break space.
Platform/font support matters.
Do not introduce unsupported characters casually.
Zero-width characters
Zero-width space/joiners can appear in scripts.
They may be linguistically meaningful.
Generic whitespace cleanup can damage text.
Language-aware tools are essential.
RTL directional marks
Invisible directional marks affect punctuation/numbers.
Do not delete unknown hidden characters in RTL content.
Use bidi-aware review.
CJK full-width punctuation
CJK typography can use full-width punctuation and language-specific marks.
Western QA can mislabel them.
Configure per locale.
French spacing
Depending on style/system, punctuation may require spaces before:
- colon;
- semicolon;
- question mark;
- exclamation.
Use target norms and available space characters.
Spanish punctuation
Opening ¿ and ¡ matter.
Do not treat extra punctuation versus English source as error.
German quotation marks
Target may use:
„…“
or project style variants.
Check outer/inner consistency.
English UK versus US quotes
House styles vary:
- single outer;
- double outer.
Follow project, not generic assumptions.
Apostrophes in names
Names such as O’Connor require correct apostrophe.
Straight/curly may matter stylistically.
Do not change spelling.
Prime symbols in measurements
5′6″ should use prime/double prime where required.
Quotation marks are visually similar but semantically different.
Mathematical symbols
Use:
- ×;
- ±;
- −;
- ≤;
- ≥.
Do not replace with approximate keyboard substitutes if formal typography required.
Trademark and registered marks
Check position after translated brand.
Do not accidentally place symbol after translated descriptor instead of brand.
Copyright year
Mechanical pass can notice:
© 2025
in 2026 document.
But updating factual year is source/content decision, not typography.
Flag rather than silently alter.
Ellipsis and sentence punctuation
Does ellipsis replace final period or coexist?
Style varies.
Use house rule.
Multiple exclamation marks
Marketing source uses !!!.
Target brand style may allow or restrict.
Do not automatically collapse if tone is intentional.
Parentheses after translation
Target syntax may reorder phrase so parenthetical content no longer sits naturally.
Mechanical pass can flag bracket balance.
Target-only review handles flow.
Brackets in citations
Academic styles use brackets structurally.
Do not convert [] to () for aesthetic reasons.
Spaces around slashes
Style can be:
and/or input / output
Use context.
Spaces around em dash
US style often no spaces; other styles differ.
Project decides.
Dashes in dialogue
Some languages use em dash to mark dialogue instead of quotes.
Literary target may legitimately add dashes absent source.
Do not flag as mismatch.
Punctuation and segmentation
A full stop can affect CAT segmentation.
Late punctuation edits generally do not re-segment current project automatically.
But future TM/source behavior may differ.
Do not change source punctuation; target punctuation is safer.
Mechanical cleanup before TM update?
If target is confirmed to TM before final punctuation cleanup, old TM entry can retain mechanical errors.
After edits, reconfirm/update working TM as workflow requires.
This preserves clean reuse.
Mechanical cleanup after monolingual review
Ideal sequence:
- target-only flow;
- punctuation/whitespace;
- QA.
Style edits can create surface errors.
Mechanical pass last catches them.
Mechanical cleanup before DTP
Fix text-side typography before layout.
DTP should not spend time correcting linguistic punctuation if CAT source can be updated.
Keep reusable asset clean.
Mechanical cleanup after DTP
Still inspect rendered output for:
- widow/orphan punctuation;
- missing glyph;
- broken no-break behavior.
Some issues only appear visually.
Search patterns worth having
Possible client/project checks:
- two or more spaces;
- space before comma;
- no space after comma;
- straight double quote;
- repeated full stop;
- hyphen between digits.
Each needs contextual review.
Build small checklist.
False positives
Example:
Two spaces intentional in monospaced code.
Do not enforce prose rule globally.
QA profile should allow scope/exceptions.
Mechanical issue comments
If unresolved typography depends on client style, comment:
STYLE—use guillemets or curly double quotes?
Resolve once.
Update style card.
Do not debate every quote occurrence.
Punctuation normalization after copy from PDF
PDF text extraction can produce:
- weird line breaks;
- nonstandard hyphens;
- missing spaces.
Source cleanup should happen before translation.
Target mechanical pass can still catch residue.
Punctuation normalization after OCR
OCR can confuse:
- apostrophe/prime;
- hyphen/dash;
- period/comma.
Verify against scan for high-risk cases.
Punctuation and locale auto-correct
Word processors may auto-replace quotes based on document language.
Ensure target language set correctly.
Otherwise external reviewer may introduce wrong quote style.
Review-package typography
Bilingual DOCX exports can alter:
- smart quotes;
- spaces.
After reimport, run mechanical pass.
Smart replacement tools
Use tools like quotation/spacing correctors only when:
- language configured;
- exclusions understood;
- review mode available.
Automation should present candidates, not destroy structure.
Whitespace after punctuation changes
Replacing semicolon with period may leave capitalization issue.
Mechanical edits can cascade.
Read local sentence boundary afterward.
Symbol font fallback
Correct Unicode symbol may display as box in final font.
Preview.
If unsupported, choose approved fallback.
Accessibility and punctuation
Screen readers interpret punctuation differently.
Do not overuse decorative symbols.
Target style should prioritize understandable output.
Plain-text systems
Some systems allow only ASCII.
Curly quotes/em dashes may be invalid.
Project technical requirements override typographic preference.
Export encoding
Even correct target character can corrupt on export if encoding wrong.
Round-trip file test catches.
Mechanical pass alone cannot.
A typography regression set
Keep representative strings:
- quote;
- dash;
- unit;
- non-breaking space;
- RTL/CJK if relevant.
After tool/settings changes, export and inspect.
Mechanical pass for marketing
More expressive punctuation.
Style-driven.
Do not over-normalize brand voice.
Mechanical pass for legal
Conservative.
Quotation/citation punctuation can have formal significance.
Use house/legal style.
Mechanical pass for technical
Focus:
- units;
- ranges;
- symbols;
- part numbers;
- warnings.
Avoid cosmetic changes to identifiers.
Mechanical pass for education
Check:
- question numbering;
- answer labels;
- punctuation instructions;
- mathematical signs.
Mechanical pass for software
Exclude:
- code;
- placeholders;
- keys.
Focus human-readable strings.
Mechanical pass for subtitles
Punctuation can affect reading pace and dialogue.
Use subtitle style guide, not document rules.
Mechanical pass for spreadsheets
Whitespace can affect lookup keys.
Do not trim structural cells indiscriminately.
Translate only reader-facing fields.
Final “show invisibles” sweep
At the end, turn on hidden characters for a sample.
Look for:
- strange leading spaces;
- double spaces;
- nonbreaking spaces;
- tabs.
Then turn off.
Use diagnostic view, not permanent clutter.
The mature typography workflow
- locale style defined;
- QA catches obvious errors;
- target-only pass handles natural punctuation;
- mechanical pass cleans surface;
- final preview checks rendering.
No one stage carries entire burden.
Final principle
Typography should disappear to the reader.
When spacing, quotes and symbols are correct, nobody notices.
When wrong, they make the translation feel unreliable even if meaning is correct.
That is why a short mechanical pass has disproportionate quality value.
Advanced practice: separate linguistic punctuation from technical punctuation
Not every punctuation mark belongs to prose.
Linguistic
- commas;
- quotation marks;
- sentence full stops;
- dialogue dashes.
Technical
- code delimiters;
- path separators;
- version dots;
- decimal marks;
- formula operators;
- markdown syntax.
A mechanical pass should normalize linguistic punctuation according to target style while protecting technical punctuation according to structure.
This classification prevents “smart cleanup” from breaking data.
Worked example 21: URL
Source/target includes:
Do not turn hyphen/dash, question mark or equals into typographic variants.
The URL is a technical token.
Worked example 22: email
name@example.com
Do not change apostrophes/periods around it.
Protect.
Worked example 23: Windows path
C:\Program Files\App
Backslashes are syntax.
Do not normalize slashes based on prose style.
Worked example 24: command flag
–verbose
Two hyphens are not an em dash.
A typography tool needs exclusions.
Worked example 25: markdown emphasis
Important
Asterisks are markup.
If CAT tool exposes them as text, do not remove as duplicate punctuation.
Know file type.
Worked example 26: JSON punctuation
{“status”:”active”}
Curly quotes would break parser.
Technical no-touch zone.
Worked example 27: mathematical interval
[0,1)
Brackets have mathematical meaning.
Do not “balance” the parenthesis.
Mechanical QA can misinterpret specialized notation.
Worked example 28: legal subsection
4(a)(ii)
Parentheses identify structure.
Do not restyle to brackets casually.
Worked example 29: bibliographic DOI
10.1000/xyz123
Punctuation is identifier.
Protect exact.
Worked example 30: ellipsis in dialogue
Source writer uses three dots intentionally for hesitation.
Target style may use ellipsis glyph.
This is linguistic.
But preserve pragmatic effect.
Spaces before/after tags
If a tag wraps word, decide where space belongs:
word <b>term</b> word
When target reorders, spacing must move too.
Tag insertion alone does not guarantee whitespace.
Space normalization and XML
Whitespace may be significant in XML preformatted fields.
Do not global trim.
File filter should protect structure, but verify.
HTML collapsing spaces
Browsers collapse ordinary spaces, but behaves differently.
Do not insert non-breaking spaces casually to align visual text.
CSS/layout should handle design.
Non-breaking hyphen in identifiers
Use to keep:
- code;
- phone prefix;
- compound.
But only if output supports.
Soft hyphen
Invisible discretionary hyphen can appear from source documents.
It may create strange search/matching.
Decide whether file pipeline should remove before translation.
Do not manually chase hundreds.
Word joiner/zero-width no-break
Some scripts use invisible control.
Generic cleanup can damage.
Language-specific expertise required.
Punctuation around inline tags
Example:
<b>Warning:</b>
Does colon belong inside bold?
Source may indicate.
Target style may differ.
Preview helps.
Punctuation around hyperlinks
Should period be inside link?
Usually link text should not accidentally include sentence punctuation unless intended.
Check.
Punctuation around quotations
Target style controls whether final punctuation falls inside/outside closing quote.
Do not copy source convention blindly.
Punctuation around parenthetical translations
If source includes term (ABC), target may reorder abbreviation.
Ensure brackets remain around correct content.
List punctuation profile
Decide for each list type:
- fragments;
- sentences;
- labels.
Then normalize.
Do not apply one punctuation rule to every bullet in entire project if list types differ.
Colon after headings
Some languages/styles use colon after inline heading.
Others not.
Use style card.
Semicolon in complex lists
Legal/technical lists may require semicolons to separate clauses.
Do not simplify to commas if ambiguity rises.
Comma splices
Target-only grammar review may catch.
Punctuation pass can flag long comma-heavy sentences.
But substantive rewrite belongs to language review.
Keep pass boundary.
Spaces after sentence punctuation
One space is common modern style.
Some client legacy style uses two.
Project rule wins.
Tabs in generated tables
Tabs may be layout controls.
If translating plain text that will be parsed, preserve.
Full-width spaces
CJK sources may contain full-width spaces.
Target behavior depends.
Do not replace automatically without locale knowledge.
Arabic punctuation
Arabic uses:
- Arabic comma
،; - question mark
؟.
Source English punctuation should not be copied mechanically.
Ensure bidi layout.
Persian punctuation
Similar but language-specific conventions.
Use target proofing tools.
Devanagari danda
Some Indic languages use danda ।.
Project/language norms matter.
Generic Western punctuation QA may be wrong.
Locale-specific quotation marks
Create mapping per language.
Example:
- French guillemets;
- German low-high;
- Polish variants.
Do not centralize one universal quote style.
Currency spacing
Examples can differ:
- $100;
- 100 €;
- CHF 100.
Punctuation/spacing pass checks style but not currency identity.
Percent spacing
Some styles:
- 25%;
- 25 %.
Locale-specific.
Ensure non-breaking space if required.
Units spacing
SI style typically uses space between value and unit, with exceptions.
Project/domain may specify.
Do not blindly mirror source.
Degree symbols
Temperature:
20 °C
Angle:
20°
Spacing differs by meaning.
Symbol role matters.
Multiplication and dimension
Dimensions:
5 × 10 cm
Use multiplication sign.
But product code:
5×10
may be identifier.
Classify.
Copyright/trademark placement
A mark may appear only at first brand occurrence by legal style.
Do not copy to every occurrence unless instructed.
Punctuation pass can check consistency with brand guide.
Ampersand in company names
“Johnson & Johnson” is identity.
Do not replace & with “and” as prose normalization.
Slash in gender-inclusive language
Some target styles use:
- slash;
- middle dot;
- paired forms.
Follow inclusivity/style policy.
Middle dot
Catalan or inclusive typography may use middle dot with linguistic function.
Do not treat as stray symbol.
Apostrophe in contractions
Straight versus curly is style.
Missing apostrophe changes spelling.
Spellcheck can help.
Apostrophe in measurements
Prime vs apostrophe, as above.
Use context.
Punctuation in filenames
Target may not translate filenames.
Characters such as colon/slash can be illegal.
File references should remain exact unless localization pipeline maps them.
Illegal characters
Some systems forbid characters.
QA can check.
Mechanical pass should respect technical constraints.
Search for suspicious Unicode
Copied text can contain:
- nonstandard hyphen;
- narrow no-break;
- smart quotes.
Not all suspicious characters are wrong.
Build client-specific allowed set if needed.
Normalization NFC/NFD
Visually identical diacritics can be encoded differently.
Most translation workflows normalize automatically.
If software identifiers depend on exact code points, engineering should handle.
Do not manually normalize blindly.
Spelling language setting
Word processors may auto-correct punctuation based on language.
Set target language correctly before final proofing.
Otherwise quote marks may flip to wrong locale.
Autocorrect risks
Autocorrect can change:
(c)→ ©;--→ em dash;- straight quotes.
Useful in prose.
Dangerous in code.
Disable or scope.
Clipboard risks
Pasting from web can import:
- NBSP;
- weird dash;
- smart quotes;
- hidden formatting.
Paste as plain text when appropriate, then apply project typography.
Mechanical cleanup and style guides
Typography rules should be explicit enough:
- quote form;
- dashes;
- spaces;
- capitalization.
If style guide silent, follow target-language professional convention.
Mechanical cleanup and client legacy files
Old files may contain inconsistent punctuation.
Do not treat them all as authority.
Use current style.
TM reuse can propagate old mechanics.
QA false positives from punctuation
Source period absent in target because label style.
Mark/ignore if legitimate.
Do not alter target merely to make dashboard green.
QA false negatives
A punctuation mark can be wrong but not differ from source.
Example source itself uses English quotes in French content.
Human target style review still needed.
Batch correction strategy
For a repeated safe defect:
- search;
- inspect samples;
- scope;
- replace;
- rerun search;
- QA.
This is the controlled-batch workflow.
Do not use global Replace All for symbols with multiple functions
Hyphen is classic.
Same document can contain:
- hyphenated word;
- range;
- minus;
- code.
Use context-aware regex if expert, or review individually.
Punctuation pass order
Efficient order:
- whitespace;
- quotes/apostrophes;
- dashes/hyphens;
- brackets;
- locale-specific punctuation;
- non-breaking chars;
- final QA.
This goes from safer to more contextual.
Why whitespace first?
Double/missing spaces can interfere with later searches.
Clean them early.
Why quotes before dashes?
Quote normalization is often more globally regular.
Dash function more ambiguous.
Process low-risk patterns first.
Final random sample
After cleanup, choose 20 random target segments.
Look only at:
- spaces;
- punctuation;
- symbols.
If no systematic defect appears, confidence rises.
Surface-quality metrics
Count:
- QA warnings before/after;
- false positives;
- recurrent symbol errors.
Use to improve QA profile.
When this pass can be skipped
Very small UI project with strong live QA and preview may not need separate typography pass.
Use proportion.
When it is essential
Long publishing documents, reports, legal text, multilingual DTP, technical manuals.
Surface inconsistency becomes visible at scale.
The mature mechanical workflow
The translator is not hand-correcting every space at delivery.
Instead:
- source clean;
- style defined;
- live QA catches basics;
- target-only pass handles language;
- focused surface pass handles mechanics;
- preview handles rendering.
Each layer does its job.
Final typography acceptance gate
Before delivery, inspect one representative page, one dense table or list, and one structurally unusual segment such as a tag-rich or code-adjacent string. Confirm that the target uses the intended quotation style, dash conventions, spacing, non-breaking characters and symbols in all three environments.
This matters because a mechanical rule that works perfectly in ordinary prose can fail around structured content. The final sample should therefore include both normal text and edge cases.
If the sample exposes a pattern rather than a one-off defect, fix the rule or QA profile and rerun the affected category. One systematic cause should be repaired once, not corrected manually in dozens of segments.
The release standard is simple: target typography should be consistent enough to disappear from the reader’s attention, while every symbol that carries technical or semantic meaning remains intact.
Summary
A whitespace, punctuation and symbol pass helps people translate quickly by batching target-side mechanical cleanup after semantic translation is stable.
The reliable workflow is:
show/scan hidden characters → run QA → fix whitespace → normalize target punctuation → verify quotes/dashes/symbols → protect meaningful special characters → preview final rendering
Do not force source punctuation onto target.
Do not replace symbols globally without context.
A clean mechanical pass makes real errors visible and the final target look intentional.
Frequently asked questions
Why run a separate punctuation pass?
Because monitoring punctuation and whitespace during semantic translation divides attention. A later focused pass is faster.
Should target punctuation match source?
Not always. Target-language and client style rules control punctuation.
What is a non-breaking space?
A space that prevents a line break between the items it joins.
Should I replace straight quotes with curly quotes?
Only according to target/client typography and content type. Straight quotes may be required in code or specific styles.
What is the difference between hyphen, en dash and em dash?
They serve different functions such as word joining, ranges and parenthetical breaks depending on style.
Can QA catch whitespace errors?
Yes. Many CAT QA systems flag double spaces, punctuation and formatting issues.
Why show hidden characters?
They reveal spaces, tabs, line breaks and non-breaking characters that are otherwise hard to diagnose.
Can punctuation errors change meaning?
Yes. Commas, minus signs, apostrophes and other symbols can alter meaning.
Should I use Replace All?
Only after searching and proving the pattern is safe across contexts.
When should I preview final output?
Whenever typography affects line wrapping, glyph rendering, UI fit or structured layout.
Internal-link opportunities
- How People Translate Quickly | QA Profiles
- How People Translate Quickly | Live QA Warnings
- How People Translate Quickly | In-Context Preview
- How People Translate Quickly | Safe Find and Replace
- How People Translate Quickly | Style Cards