We are now getting into some very practical applications for metadata. Of course, the most important part of your strategy is to make the metadata useful, which means that it needs to be discoverable. In this post, I’ll outline some of the ways that metadata can be displayed to enhance user discovery. We will also take a look at some of the ways we are building these discovery tools into the next version of our software: Tandem Vault 3 (TV3).
As you think about how to structure your metadata, the method of presentation can be an important consideration. Before we discuss the metadata planning process, let’s take a look at the way that metadata can be presented to users of the collection. Each type of information lends itself to a specific type of interface, and understanding the strengths of each interface should be part of your taxonomy development.
Note that the list of interface options shown below will not exist in every piece of software. Your choice of software and services should be informed by the structure that best supports the kind of information your collection and your stakeholders require.
What’s wrong with a search box?
In the Google age, we have come to expect that our discovery needs can be met with a simple search box. That’s not always the case, and it’s seldom the case for specialized collections (like the ones a company or organization has.) The Google search bar sits on top of a lot of sophisticated data processing which your collection probably does not have.
There is a ton of logic, relationship graphs, and natural language processing sitting underneath the surface which is the secret sauce for Google’s success.
Google also knows a lot about you and what you might be looking for. It can factor in past searches, both yours and searches made by others. Private collections generally don’t have access to the data and algorithms necessary to do a Google-like search of their own collection.
Without the backend logic, blind searches generally become frustrating and unproductive. That picture of the cow you want may be sitting in the collection, but you searched on Cattle and got no results. How are you supposed to know which tag to search on? The collection manager may know, but that does not mean the general users know.
Let’s look at some of the ways metadata can be presented to assist in discovery.
Directories are a great tool for presenting the contents of a collection. If the hierarchy of the directory makes sense, people intuitively know how to click through folders and browse. However, “real” filesystem folders are often problematic for this, since any file can only be in one folder at a time, and it’s common for files to be classifiable in multiple ways. So instead of a directory tree that behaves like a folder, we want something more flexible.
Virtual folders or virtual groups have an advantage over conventional folders because an object can be “in” multiple groups at the same time. So you could make a browsable set of groups organized by department, as well as a set of groups organized by person, location or activity. When designed properly, virtual groups are the most valuable tool for making a collection navigable by large groups of people.
Virtual sets are really curated groups, assembled in order to communicate the content of your collection. They should be created according to your priorities for the media. Of course, building a navigable set of groups takes time, thought, testing and revision. But it’s also the easiest way to encourage other people to find what they need quickly.
Here’s a sneak peek from the Tandem Vault 3 (TV3) beta. A browsable hierarchy like this can help your stakeholders learn about your company, not just find the right file. In this case, the hierarchy tells the user the types of work the company does, and the names of some of the important projects. This information, all by itself, can help people understand both the collection and the company itself.
Lists are great for tag categories that have a limited number of variables to choose from. Because the values are shown in the list, it helps people find what they are looking for, even when they are not familiar with the collection. Short lists may be represented as a field with some tags in it, as a pulldown, or a checkbox tree (useful when you might want to add multiple values). When lists get too long, they become difficult to use.
Short lists are important for focusing attention on a particular type of information. They are also very useful for tagging, since they create a targeted request for annotations.
Here’s another shot from TV3 beta. The “Tag Suggester” is a way to create a targeted request during the upload process. This information becomes a filter when someone is browsing or searching the collection. Note that small lists can provide extra context beyond the name of the tag itself. In the case shown above, the tag is directly related to a trade show experience: you know that “cool stuff” is not just random cool stuff, but is “cool stuff seen at trade shows.”
Just like with a folder or virtual folder tree, a hierarchical structure for your metadata tags helps people to navigate the information and better understand the content of your collection. Organizing your tags into a hierarchy allows you to group similar items together, which is a lot friendlier than a gigantic list in alphabetical order. Airplane, Helicopter and Jet can all sit next to each other inside Aviation instead of separated by dozens or hundreds of unrelated tags.
Sometimes the information fits really neatly into a hierarchy, such as with the IPTC location tags. These are structured as Country>State>City>Sublocation. While this structure does not work for absolutely everything, it can cover the vast majority of your collection.
Other information may not fit so perfectly into a hierarchy. For instance, if you divide people into groups, friends, family and colleagues, how do you handle people who cross categories? To keep yourself sane, it’s probably best to just make a good effort, and not worry about absolute perfection. This is especially true if the taxonomy can be easily searched, so that you can find that person’s name, without entirely depending on the hierarchy. And some tags may stored in a catch-all category.
Again, one of the main benefits of a hierarchy is to tell the casual user what kind of content is in the collection.
Here’s a shot of the taxonomy from TV3 (which may look kind of familiar to Lightroom users). Keywords can be arranged so that they fold up nicely, but can expand into as many subgroups as you need. It provides an intuitive way to browse. We also tell users which group selected images belong to as another method for telling users about the content itself.
Tag clouds are a space-efficient way to tell the user what’s in the collection. Frequently, the tag clouds will include a count of the instance of a tag, and present that in descending order from most common to least common.
Tag clouds aren’t great as a primary means of navigation, because they become hard to use once you have more than a couple dozen results. But they are quite useful to give a quick overview of secondary information. Select a group of files and a tag cloud can quickly tell you about the content of that set of files. By gathering up the secondary terms, and showing them in order of frequency, you can quickly scan the tag cloud to get a better understanding of the content of shown items.
Here’s the TV3 tag cloud. It tells you about the group of items you’re viewing in a more consolidated way. Of course, each of these tags can be clicked to run as a filter on the set to narrow down to just those items.
And after all that, sometimes search
All of the discovery methods listed above will be more helpful to your stakeholders than a simple search, because they educate the user about the collection. Sometimes a regular search must be used though, because the information only exists in deeper fields like a caption or a full text search of a document.
It’s becoming more common for search to tell you the class of your results, rather than showing you a list. We really like search that includes information about the source of the results, so that you can more quickly find the thing you are looking for.
Also useful for metadata creation
Some of the interface items described above are also really useful for applying metadata. Placing files into a virtual group hierarchy is an act of annotation because you are recording information about the content. Likewise, pulldown lists are great for applying important tags consistently. And placing files into a keyword taxonomy automatically assigns the entire parent tree to the file.
In order for this to work, of course, the metadata element needs to be capable of two-way operation–both tagging and filtering.
Here’s the TV3 “simple” search box. We let you cast a very wide net (Search All), or narrow your search down to a particular field right from the search box.
Putting it all together
Making large collections accessible to many stakeholders requires the use of all the tools listed above. In Tandem Vault 3, we’re building a framework to make these tools easy to understand, configure and use. Keep an eye open for more information soon, or sign up to be notified when TV3 beta becomes available: https://www.tandemvault.com/stay-informed/.