Metadata
The symptom of a missing convention
You type port into the keyword field and the tool suggests: portrait, Portrait, portraits, studio portrait, Studio-portrait, port, Porto. Seven entries for three notions.
None of those values is wrong. The problem is that none is the value: there’s no convention, so every entry invents one.
The five rules that hold a list together
1. One form per notion. Pick singular or plural, lowercase or capitalised, and never deviate. portrait, not Portraits. The rule matters less than its consistency.
2. Short, orthogonal tags. A tag describes one dimension. studio-portrait-black-and-white mixes three: better portrait + studio + black and white, freely combinable.
3. No tag for something that is already data. Year, body, lens, project already live in metadata or structure. Duplicating them as tags creates two truths that will diverge.
4. A tag you won’t filter on is useless. Before creating one, ask: will I ever search for exactly this? If the tag will only ever apply to three images, a caption is enough.
5. Thirty to fifty tags, not three hundred. Beyond that you no longer remember the list, so you reinvent entries, so the list grows. It’s a vicious circle that starts around a hundred.
Three axes are usually enough
- Subject: what the image shows (
portrait,landscape,architecture,protest). - Nature of the work: production context (
commission,personal,reportage,test). - Technique, if it’s relevant to you (
film,black and white,long exposure).
Places, people, and dates are not tags: they’re dedicated fields, normalised as entities. Putting them in tags loses the structure and rules out any serious geographic or chronological search.
Catching up an existing collection
Don’t start by writing the ideal convention. Start by measuring: list your keywords by descending frequency. The top of the list is your real vocabulary. The long tail at one or two occurrences is noise to merge or delete.
Then use aliases rather than replacements: Portraits becomes a synonym of portrait, and both find the same images without you re-editing 4,000 records.
In practice
- List your tags by frequency: the top is your vocabulary, the tail is the work.
- Fix one form rule and write it down somewhere.
- Split compound tags into orthogonal dimensions.
- Move places, people, and dates out of tags into dedicated fields.
- Merge with aliases, not destructive bulk re-editing.
- Deliberately cap the list, and review it once a year.
Obscura Flow supports synonyms and aliases to merge without destroying, keeps tags and thesaurus separate from place and person fields, and offers bulk editing to catch up a vocabulary that has drifted.