How to Normalize Title Capitalization Without Losing Meaning
Safe capitalisation cleanup protects official styling, names, acronyms, scripts, numerals, and edition markers before applying any general casing rule.
In short: Do not apply a universal title-case function to the whole catalogue. First identify each title’s language and protected tokens—official styling, proper names, acronyms, numerals, scripts, punctuation, and edition markers. Apply a documented rule only to ordinary words, preview every change, manually review exceptions, and test search and identity before approving a reversible batch.
Capitalisation looks cosmetic, yet it can distinguish a brand, acronym, stylised work, or language-specific convention. A naive rule can turn a recognisable title into an inaccurate one.
Decide the purpose of normalisation
State the problem precisely: inconsistent all-caps imports, random lowercase records, duplicated styles across sources, or household-created labels. Do not rewrite a title merely because another style is preferred.
Preserve the work’s identity and the source’s authoritative form where reliable. If multiple records use different titles for the same work, resolve identity and language before capitalisation.
Build the protected-token sheet
Split each candidate title into meaningful components:
| Full title | Language/script | Token | Type | Protected form | Evidence | Decision |
|---|---|---|---|---|---|---|
| ordinary/proper/acronym/numeral/styled/edition |
Protect:
- proper names and intentional internal capitals;
- acronyms and initialisms;
- Roman numerals and numbered instalments;
- punctuation that changes meaning;
- stylised official names supported by evidence;
- words in scripts without case distinctions;
- edition, part, or episode markers governed separately;
- articles and particles whose treatment depends on language.
The sheet creates an exception register that can be reused instead of rediscovering the same title every cleanup cycle.
Write language-aware casing rules
Choose a convention for ordinary words in a defined scope: sentence case, source-authoritative casing, or a documented title style. Record how the rule handles the first and last word, short function words, hyphenated compounds, apostrophes, subtitles after punctuation, and mixed-language titles.
Do not infer a title’s language solely from the interface language. If language is unknown, leave the title for manual review.
Dublin Core defines “title” as a name given to a resource; normalisation should continue to identify that resource, not impose uniform typography at the expense of meaning.
Generate a preview, not an immediate overwrite
For every candidate, show:
- current title;
- proposed title;
- changed tokens highlighted;
- rule and exception applied;
- evidence source;
- likely search aliases;
- reviewer decision.
Reject transformations that alter punctuation, diacritics, numerals, script, or official styling outside the approved rule. Keep the original value in the change record and preserve a rollback route.
Pilot a varied batch
Include ordinary English titles, proper names, acronyms, numerals, subtitles, hyphenated words, multilingual titles, and a deliberately protected style. Then:
- apply changes through supported controls;
- refresh normally;
- search by current and familiar former forms;
- compare versions and similarly named works;
- inspect web, mobile, and TV presentation where supported;
- verify no source refresh restores a competing form;
- revert any ambiguous result.
Use the metadata audit to record severity and provenance. For category labels rather than work titles, follow the category-name standardisation guide.
Handle source conflicts
When two connected sources disagree, retain provenance and determine which value is authoritative for the specific record. Do not silently blend a title from one source with edition metadata from another.
Norva can organise compatible authorised sources, while source metadata may vary or refresh independently. A local display correction may therefore require a durable mapping rather than a one-time edit.
Common mistakes and limitations
- Applying one English rule to every language.
- Lowercasing acronyms and names.
- Removing diacritics while changing case.
- Treating all-caps artwork as title evidence.
- Rewriting subtitles and edition markers indiscriminately.
- Discarding original values and provenance.
Capitalisation checks cannot resolve conflicting identities or release years. Use the release-year investigation when titles differ because records may represent different works.
Frequently asked questions
Should all-caps official titles remain all caps?
Retain intentional styling when reliable evidence and household readability support it. If the source provides only an all-caps export convention, distinguish that from official styling.
Can search aliases preserve old casing?
Case-insensitive search may already do so, but do not assume. Test familiar forms and record whether an explicit alias is supported or necessary.
Should episode titles use the same rule as series titles?
They can share a base convention, but episode markers, quoted titles, and source hierarchies may need separate protected-token rules.