How does PhotoStructure extract keywords from my photos and videos?

April 21, 2020

PhotoStructure uses a bunch of methods to pull out keywords from your photos and videos.

Keywords from metadata 🔗

PhotoStructure extracts keywords from the following metadata tags:

CatalogSets
Categories (this is typically XML-encoded)
HierarchicalSubject
Keywords
LastKeywordXMP
Subject
TagsList
XPKeywords (these are keywords added by the Windows Explorer)

Metadata keyword extraction can be disabled with the tagKeywordsFromMetadata library setting.

Keywords from sidecars 🔗

PhotoStructure looks for metadata in .XMP, .MIE, and Google Takeout .json sidecars. PhotoStructure supports both full name and base name sidecar names. Matching is done case-insensitively.

As an example: IMG_123.JPG will inherit tags from both IMG_123.JPG.XMP and IMG_123.XMP.

Newer files will overwrite the values from older files, so by editing a sidecar, that value will “win.”

Keywords from path names 🔗

Tags found in filenames or parent directories that follow -- are added automatically. Here are a couple examples:

All files found in the /Users/bob/Pictures/2019-02-14/--event travel/ directory would be tagged with Keyword/event and Keyword/travel.
The file /home/karen/2018-11-23/P317812--ocean.jpg would be tagged with Keyword/ocean.
Filename tags can be hierarchical by using the keywordPathSeparators setting: for example, /home/alice/Pictures/--event|wedding,people/example.jpg would be tagged with Keyword/event/wedding and Keyword/people.

Path name keyword extraction can be disabled with the tagKeywordsFromPath library setting.

Multiple keywords 🔗

PhotoStructure splits keywords by commas and semicolons. For example, car, blue, tree will be interpreted as having the keywords car, blue, and tree.

This is configurable via the keywordDelimiters library setting.

Hierarchical keywords 🔗

All of PhotoStructure’s tags are “hierarchical,” rather than “flat.” Read more about PhotoStructure’s hierarchical tags.

PhotoStructure looks for “path separator” characters in keywords. By default, the characters /,|,>,≻, and ⊃ are considered “path separators.” This set can be changed by the keywordPathSeparators library setting.

This allows for tags like Family|Einstein|Albert, Flora ⊃ Fruit ⊃ Orange, or Fauna > Oceanic > Pelican.

Where do these keywords go? 🔗

PhotoStructure looks for one or more path separators in a keyword. If none are found, the keyword is “flat.” If it finds one or more of these characters, it considers the keyword to be “hierarchical.”

Non-hierarchical, “flat” keywords are added to the Keywords root tag (which you can visit via the navigation menu). As an example, an image with the keywords “blue” and “sky” will result in the asset being tagged with Keywords/blue and Keywords/sky.

Hierarchical keywords are added directly. As an example, an image tagged with Family|Einstein|Albert will introduce a new Family “root tag.”