As learning professionals, we strive to find the best way to store and deliver content, seeking out tools and processes that will save time while increasing the value of each minute of learning. To accomplish this, we need a true single-source method that would enable learning developers to author content in one place and publish into the various materials they will share with learners. In addition, the ideal solution would be able to interpret and make connections between different pieces of content. 

Enter Intelligent Structured Content. 

Before we dive into defining Intelligent Structured Content, let’s take a step back and look at its components—three individual terms that are more easily defined on its own. 

• Content: This is the message in your work. Defined in this context, it is the text, titles, images, and media elements that tell your story. This is the component that developers spend the most energy perfecting, as it is the purpose for the other two components. 

• Structured: Simply put, this refers to having clearly defined structure and organization. In other words, it’s a collection of parts organized in a specific way. 

• Intelligent: In this context, the intelligence is acquired from the structure. The content becomes “intelligent” by virtue of how it is structured and how the structural elements tie content together and add meaning to the relationships between content parts. For instance, an additional layer of structure is how it organizes content not just by “what it is,” but also by “how it will be used or consumed.” This level of intelligence allows parts of the content to be essentially self-aware. 

With simple definitions out of the way, let’s now take a look at putting the terms back together. 

Structured Content 

In essence, “structured content” refers to breaking down content into the smallest definable parts and putting it back together using a structure to organize it. We do this by reducing the content to data in its simplest form and separating it from anything that speaks to how it will “look” or “behave.” 

Once we strip away the layers of formatting, style, and interactivity, we are left with data. Then, we can structure that data into elements that define it. We call that definition a “type.” 

Some of the benefits of working with Structured Content include: 

• We can publish content data into a number of different formats with endless possibilities of formatting and style—without affecting the content.

• We can publish our content data into many different formats without duplicating the content.

• To publish content, we only need to build a list of instructions to convert each type of content into a formatted structure, we do not need to format each piece individually.

• Because we can make changes to the look-and-feel separately from the content, we can reduce time for testing content. Also, we only have to make changes to a type of content in a single place, and all the occurrences of that type will behave accordingly.

• Content, as data, can be easily translated into other languages, updated, and modified without affecting published versions. This enables effective single sourcing. 

There are also some disadvantages of Structured Content: 

• Specific skills are required to implement a structured content system, especially the engine that “publishes” the data into a usable file type. There are tools available, but they require a learning curve. The alternative is a custom-built system to manage content and publishing.

• Visual changes to how a type of content is displayed are not individual, they apply to all occurrences. This may present formatting limitations to those who prefer to fine-tune each piece of content.

• There is generally a larger front-end effort in a structured content project. Effort up-front is needed to decide on the structure and the published formats, as well as the system design and review process. All of these steps need to be considered beforehand, as they are more difficult to change once the system and process are in place. 

In the following examples, we will use simple XML to illustrate structured content, because it is relatively easy for humans to read. 

Example 1: A Piece of Content 

**API

**An Application Programming Interface (API) is a way for a computer application to expose certain parts of its functionality to other computer applications. 

In its simplest form this content is a heading (in bold) and a paragraph. In a structured content document, the heading would be a type and a paragraph would be another type, because they behave differently. Using different types to define content as data will allow them to behave differently when they are viewed by a learner. 

Example 2: Structured Content 

<heading>API</heading>

<paragraph>An Application Programming Interface (API) is a way for a computer application to expose certain parts of it’s functionality to other computer applications.</paragraph>

The content—now in data form—does not have formatting, but it does have structure. The different parts are defined; when these parts are styled for viewing by a learner, they can be formatted differently. If we wanted to format this data as it appeared in example 1, we can tell our content that “heading type” data is to be presented as bold text followed by a line break, and we can format the paragraph type content as normal weight and left aligned block text.

These pieces of data can also have relationships to one-and-other, which can be achieved by structuring both elements into a third type.

Example 3: Relationships 

<item>

API

An Application Programming Interface (API) is a way for a computer application to expose certain parts of it’s functionality to other computer applications.

</item> 

If our content is defined and those relationships are made, we can change the way a user consumes the content without changing the content. In this example, we have built a relationship between the heading and paragraph, as they are both products of an item. We can now define the behavior of item in a unique way, even though it is comprised of fairly generic elements. Another way to look at this is that a machine (computer system) can uniquely identify our heading as it relates to the paragraph within this item.

Intelligent Structured Content

Intelligent structured content is an extension on structured content. To make structured content intelligent, we need to make it self-aware. We need to extend the definition of an element beyond what it is (type/behavior) to what it means (content/context). 

In Example 3, we had an item that was comprised of a heading and a paragraph; these elements are structured but do not have awareness of the content they contain. Intelligent structured content is the practice of defining those elements to add context to the content they would contain. If we define the item as a component of a glossary and redefine the heading as a glossary term then further define the paragraph as the glossary definition, we have added a layer of intelligence. 

Example 4: Added Intelligence in the Structure 

<glossary>

<term>

API

</term>

<definition>

An Application Programming Interface (API) is a way for a computer application to expose certain parts of it’s functionality to other computer applications.

</definition>

</glossary> 

We can still publish this content to look and behave in the same way structured content would, but our content has become self-aware. The heading is now aware that it is a term in an item of a glossary. Much like structure, this intelligence can be integrated from the granular elements right up to the whole document. This adds stronger more contextual relationships to the content parts. Now, machines can determine not just how the content should behave, but also what the content means. 

How Do We Use the Intelligence? 

All of the advantages of structured content apply to intelligent structured content. However, the interesting added advantage of intelligent structured content is that it allows the world of machines to “read” the content. Allowing a machine to read the content—and not only its structure—gives the machines the ability to work with that content in context. Making context available to the machines that are storing and sorting content is powerful and opens up a world of possibilities as our content libraries become increasingly dense. 

Possible Advantage 1: Searching. Search engines are the standard starting point for us to find the content we need. I say “starting point” because we are ultimately responsible for sifting and filtering the results down to what we are actually seeking. This sift-and-filter takes time. Even when browsing through a taxonomy or index of topics, time and cognitive effort is required while we narrow down our search from many options to a select few. 

At its root, this process is a single task: “Is this what I’m looking for?” yes/no, repeat. 

The fewer results we have, the faster we’ll find something useful. Machines can perform the same task repeatedly in a shorter time with greater consistency than we can. The missing link is context, though. Generally, machines are only looking for a keyword or a pre-defined tag to gather results for us. However, typically that grouping of potential content lacks context. 

Intelligent content enables machines to search within specific context. This makes sifting and filtering very efficient as many of the pieces returned already fit the context that we are searching within. With added intelligence, a machine can target a certain type of content. 

For example, if we asked a machine to look for glossary entries related to computer science, it could narrow the search to “glossary” type structures and build a result list that is more accurate than a simple keyword search on all content for the same term. Even with metadata that tags the content as containing the term, it may not occur in context of a glossary. 

Possible Advantage 2: Content Filters. The added intelligence in the structure can also allow machines to filter documents for specific types of content. Machines can easily extract and compile all glossary terms and present those in a separate document to be published on its own. This same filtering can be applied to larger structures to create published libraries of very specific content types. 

The opposite is also true, and certain pieces of content or whole types can be excluded from being published. If the content contains intelligent structures for instructor notes or question answers, they can be excluded from the “learner” published package but included in the “instructor” package. This filtering also can apply to different languages contained in the same structure and content variations catered to different publishing media (web vs. mobile vs. print)—effectively allowing a true single source document for content. 

Possible Advantage 3: Contextual Recommendations. We have covered that machines are good at sifting and filtering content. If we put both of those together and use the context of who we are (our user profile) and what content we are experiencing, we can let machines suggest where to go next. 

Extend that idea to perhaps where we are (GPS coordinates) and what we’re doing (the system we are interacting with) and machines could push content to us if they observe that we are having difficulty. Our intelligent content contains enough information that the machines, with help from human teachings, can start to make suggestions based on context. That context makes those suggestions relevant and meaningful. 

Bottom Line 

Intelligent structured content is a method to structure data to define meaning to the content. This method enables content to be self-aware and intelligent. More importantly, intelligent structured content allows machines to determine the context of the content, as well as the structure. This allows machines to understand what content means to the users and developers. 

If we provide the rules, machines can make very accurate recommendations—based on how we want to use the content. The boundaries of how accurate those suggestions can be are still unknown, but we do know that it all starts with intelligent structured content. Perhaps with enough time and testing, we could let the machines aggregate entire learning paths for us!