How to Normalize Names & Tags

Audience: Archivists — people cleaning up inconsistent names, tags, and locations extracted from contributor responses.

When multiple contributors describe the same photos, they naturally use different phrasings for the same people, places, and tags. One person says "Bob," another says "Uncle Bob," and a third says "Robert Johnson." Entity normalization lets you tell Zeuge that all three refer to the same person, so your archive uses a consistent, canonical name.

Why Normalization Matters

Without normalization, searching or filtering by "Robert Johnson" would miss all items tagged "Bob" or "Uncle Bob." Normalization ties all these variants to a single canonical form, making your archive much more useful as a searchable record.

What Gets Normalized?

Zeuge supports normalization for three entity types:

Type	Example Variants	Canonical Form
People	"Bob," "Uncle Bob," "Bob Johnson," "R. Johnson"	"Robert Johnson"
Tags	"beach house," "beachhouse," "the beach"	"beach"
Locations	"Maine," "Maine beach," "Maine USA"	"Maine"

Step 1 — Open the Metadata Tab

From your batch detail page, click the "Metadata" tab. The normalization tools are available in this view.

📸 [SCREENSHOT: Metadata tab with the normalization panel or button visible]

Step 2 — Get Normalization Suggestions

Click "Get suggestions" (or "Suggest normalizations"). Zeuge will analyze all the names, tags, and locations extracted from every response in the batch and find groups of strings that look like they might be the same entity.

📸 [SCREENSHOT: The normalization section — before and after clicking "Get suggestions"]

The suggestion algorithm works by:

Collecting every people name, tag, and location from all responses.
Using fuzzy string matching to group similar strings together (e.g., "Bob" and "Bob Johnson" are close; "Bob" and "Carol" are not).
Suggesting the longest or most complete version as the canonical form.
Showing how many items are affected by each suggested merge.

Step 3 — Review Suggestions

The Normalization Modal opens, showing a list of suggestion cards. Each card shows:

Group members — all the variants found (e.g., "Bob," "Uncle Bob," "Bob Johnson")
Suggested canonical — the form Zeuge recommends (e.g., "Bob Johnson")
Items affected — how many items would be updated if you merge

📸 [SCREENSHOT: The Normalization Modal showing several suggestion cards with group members, suggested canonical, and affected item count]

Step 4 — Accept, Customize, or Skip Each Suggestion

For each suggestion, you have three options:

✅ Accept — Merge with Suggested Canonical

Click "Merge" to accept the suggestion as-is. All variants in the group will be unified under the suggested canonical name.

✏️ Customize — Change the Canonical Name

If you disagree with the suggested canonical, type your preferred form in the "Canonical name" field before clicking "Merge."

Example: Zeuge suggests "Bob Johnson" as canonical, but you know his full name is "Robert Johnson." Type "Robert Johnson" and click "Merge."

⏭️ Skip — Ignore This Suggestion

Click "Skip" if the grouping is wrong — e.g., Zeuge grouped "Carol Johnson" and "Carol Smith" as the same person, but they're different people. Skipping leaves both values as-is.

📸 [SCREENSHOT: A single suggestion card showing the "Merge," "Skip," and the custom canonical name input]

Step 5 — Save Your Normalizations

After reviewing all suggestions, click "Save" (or the equivalent confirm button) to apply all your accepted merges.

Zeuge stores a Batch Normalization document that maps:

Each variant (alias) → its canonical form

This mapping is used when querying and filtering the archive's metadata.

📸 [SCREENSHOT: The normalization saved state — showing the list of canonical entities with their aliases]

Viewing & Managing Existing Normalizations

You can view the current normalization state at any time from the Metadata tab. The normalization section shows:

Each canonical name and its known aliases
When the mapping was last updated

You can manually edit the aliases list if needed — for example, to add a variant that the suggestion algorithm missed.

📸 [SCREENSHOT: The normalization entity list showing canonical names and aliases for people, tags, and locations]

Running Normalization Again

You can run the suggestion process multiple times as new contributor responses come in. Each run re-analyzes all current responses. Previously saved normalizations are preserved — Zeuge won't re-suggest merges you've already accepted.

Common Questions

Will accepting a suggestion automatically update the metadata on every item? The normalization mapping is saved and applied when querying or exporting your archive data. It functions as a lookup table — the raw extracted values in each response are not edited, but the canonical form is used wherever a normalized view is needed.

What if two variants shouldn't be merged — they're different people? Click "Skip" on that suggestion. If Zeuge grouped them in error, skipping prevents the incorrect merge.

My aunt goes by three different names depending on which contributor you ask. Can I normalize all three? Yes. If all three appear in the same suggestion group, merge them to one canonical. If they appear in separate groups, you can merge each group to the same canonical name.

Can I normalize across multiple batches? Currently, normalization is per-batch. If you want consistent naming across your entire archive, you'll need to run normalization in each batch separately.