Blog / Foundations, Metadata

Describing Media with Keywords

Peter Krogh
Thu Sep 10 2020

Keywords are words or phrases that you associate with a picture to describe the subject matter, style, or connotations of the image. These descriptions can be of great use when organizing, filtering, and searching your media collection.

Keywords can be abstract terms (like “victory”) or subject-oriented terms (like “cat” or “Maddy”). Subject-oriented terms are generally easier to apply because they require less careful consideration. Abstract terms are generally economical to apply only to the very best images, such as your highest rated ones or ones that will be made available in a searchable stock photography database. Both of these types of keywords can be created by AI services with varying degrees of success.

What’s a Keyword?

The term “keyword” can have two meanings: it can refer to the descriptive term itself, or it can refer to the place where the keyword might live: the IPTC keywords field (which is, technically, The Dublin Core Subject field). It’s important to understand this difference — just because you want to associate a term with an image does not necessarily mean that you want that term to live in the keywords namespace.

When keyword-type information fits in one of the other supported schemas, it’s usually best to put it in the defined namespace (e.g., Location). Putting the term in a defined namespace can remove ambiguity from the tag. Tags for photos taken in Intercourse, Pennsylvania are less confusing if the word “Intercourse” is sitting in the City field rather than the keywords field. Of course, if the picture is about the location instead of just taken in the location, you should consider adding it to the keywords field as well.

Photo by Luke Wisely Licensed as

Hierarchical Keywords

Keywords are a wonderfully flexible tool to describe what an image may be about, but they can get unwieldy on a collection-wide basis. A simple keyword list can easily climb to hundreds or thousands of terms and thus become difficult to navigate. Since most programs display keywords alphabetically, synonyms and alternate terms may not appear anywhere near each other. “Airplane,” “Helicopter,” “Jet,” and “Plane” are all terms for aircraft, but they will be widely separated in an alphabetical keyword list.

This problem is being addressed by the Adobe Hierarchical Subject field which defines relationships between keywords, and allows them to be grouped into hierarchies. So instead of “Helicopter” and “Airplane” being part of a single flat list, they can both live inside the parent term “Aircraft.” “Aircraft” could then live inside “Vehicles” or “Transportation” (or both). Using the hierarchy makes it easy to navigate your keyword list and to keep related terms close to each other.

The hierarchical keywords field, as it is currently designed, accomplishes two things. First, it offers a way to organize the keyword display to make the information easier to navigate (“Aircraft” is a child of “Vehicles”). This function is primarily oriented toward making a keyword list more user-friendly. Second, the hierarchy can provide context for a particular term (this “Madeline” is a child of “Family>Krogh” rather than “Friends>Bankson”). This function is about making the context of the keyword more portable between applications.

Note that while the older dc:subject field is widely supported, many programs don’t see the hierarchy field at all. Even when a program sees the two fields, it may not handle them properly since there is no clear policy on how to handle the synchronization between flat and hierarchical fields. There is also no good way to reconcile similar hierarchies that may be missing terms.

For this reason, it’s best to consider keyword hierarchy to primarily be a tool to use inside a single collection, and to not expect it to translate smoothly between multiple applications or even multiple catalogs that are made by the same application.

Using Hierarchical keywords

If you choose to use hierarchical keywords to organize your collection, keep a couple of things in mind:

  • Be as consistent as you can with hierarchies within your own collection.
  • The hierarchy will probably not translate directly from your collection to anyone else’s.
  • You may need to migrate the data structure to another tool eventually.
  • When in doubt, the flat keyword list is the most universal.

It’s quite possible that AI/ML tools will do a better job solving the problem that Hierarchical Keywords was created to address. That problem, “what does this keyword mean, and how do I manage a long list of keywords” is certainly an issue that AI tools are trying to fix.

Making and using keywords

As an all-purpose place to store lots of different information, it’s likely that your keywords will grow in an organic manner. As your collection grows there are all kinds of new things you might want to tag, and keywords are a great bucket to collect these tags in. Here are some tips for making use of keywords.

Make keywords for what’s important to you

I’ve run across people who think that they need to add dozens of keywords to an image to do it “properly.” That might be the case for a stock library, but for a lot of collections, a few keywords on any image may be perfectly adequate. Keyword for terms that are relevant to you, your collection, and the people who use it.

Keyword organically

I suggest you approach keywording on an organic basis, doing the proper amount of work at the proper time. What do I mean by that? During initial ingest, it’s the best time to add some basic keywords to identify a shoot or group of images. It also may be appropriate to add some other particularly relevant keywords, but you probably don’t want to spend a lot of time adding everything you can know about an image to each picture.

Additional organic keywording can happen as projects or image requests arise. Perhaps you need to make a collection for the CEO’s retirement, best facility photos or dramatic clouds to composite into another photo. While all the CEO, facility or cloud photos may not have been keyworded as you went along, you can do it now. Any time you are doing a comprehensive search through the collection for something, you should probably be keywording the results. This leverages your work and gradually adds intelligence to the media library.

Add keywords for assignments and events

Pictures that are made for a particular event or for a particular assignment should generally be keyworded for that shoot. This could be the name of the event or assignment. There are typically a handful of terms that can be applied to the entire set.

When should another tag also be a keyword?

As we have learned, there are a lot of metadata fields dedicated to certain types of information like people, locations, products and events. Does this mean they should not appear as keywords? Not necessarily. Here’s how I suggest you approach this issue.

First off, keywords are a pretty accessible field. As we saw in a previous post, keywords are indexed by many programs and even your operating system itself. This makes it a very useful and visible place to put relevant information. If you are going to send photos to people who may not have access to the deeper metadata fields, adding the information as keywords may really help them out.

If the tag is really important to the photo, it’s probably good to write it in the keywords field and the dedicated field. Not every picture taken in New York is about New York, but when it is about the location, then it’s appropriate to add the keyword.

Preserving knowledge

And finally, there is another category of information that may also inform your keyword strategy. If you are using your media collection as a memory tool, then you may want to be more expansive in your use of keywords. Your media collection is a great place to store institutional knowledge: personal, family, or corporate. If you are thinking about your media collection in this way, it makes sense to add even more keywords.

Of course, some important institutional knowledge can fit in a keyword or short keyphrase, but sometimes it takes a whole sentence or paragraph to communicate the relevant information. In this case, the caption is probably a better field.

Is Mediagraph right for your organization?

Let’s find out together.

Book Your demo today