Is Rich Narrative Possible From Structured Content?

Two of the biggest themes in content these days are structured content, and storytelling.  A number of people are suggesting the two approaches complement each other.  For example, Robert Rose, a well-known promoter of content marketing, commented in a LinkedIn discussion: “It’s not just about tags, taxonomies and 1’s and the 0’s… And it’s not just about the storytelling, personas and buyer’s journey… It’s where those things — and most importantly people —  meet that move the business forward.”  The proposed combination of structured content with narrative has sparked both anticipation and uncertainties.  Will it be transformational and game-changing, or a hornet’s nest of anguish?

Storytelling and factually oriented technical communication can potentially learn from each other.   The structured content approaches associated with technical documentation enable content to scale up for use by many different people in different channels.  The storytelling approaches embraced by the best content marketing and journalism, when backed by robust analytics, can enable content to be genuinely wanted by people, rather than just be minimally adequate.

Even if we have a solid rationale for trying to get the two concepts to work together, that doesn’t guarantee the effort will be successful or easy.  How editorial structure (the framework behind storytelling) and content structure (the framework behind intelligent content) work together is still unclear.  The two approaches are different in their origin and purpose, and anyone curious about how they might complement each other needs to understand their differences.

So far, there are few concrete examples of semantic content use in narratives.  Many different kinds of issues are involved: technical, emotional, and pragmatic.  It is hard to separate what’s visionary potential from what’s wishful thinking.

The potential of semantic content is linked to a number of foundational questions. These include:

• What is the relationship between editorial structure, and the structure of content as understood by computers?
• How far can stories be structured?  To what extent does structured content support narrative?
• How does structure support or hinder communication — the ability of people to understand on their own terms?
• Is structure the same as modular reuse?
• Can the modular content techniques developed for technical support documentation be readily applied to narrative-driven marketing collateral?

Besides presenting an intriguing goal, the topic holds larger significance.  The proposition that semantic content and storytelling can be combined challenges the discipline of content strategy to examine its assumptions, and consider possibilities for innovation.

The Many Dimensions (and Guises) of Intelligent Content

Intelligent content is a term used to describe approaches for making content more intelligent to both audiences and the computers serving them.  It is an umbrella term covering a range of different related concepts: structured content, modular content, atomic content and reusable content.  Because there is significant overlap in these concepts, there can be a tendency to treat them as equivalent.  On occasion we use these terms interchangeably, which in some contexts is appropriate.  At the same time, there can be differences in meaning and nuance among these terms we should be aware of.  As best I can tell, there is no consensus in the content strategy community defining these terms, and as a result, they are sometimes used in somewhat different ways.  To me, the terms can suggest slightly different things:

• Semantically structured content — how the structure of an article or episode affects its meaning, expressed through machine-readable metadata such as page section descriptions.  For some people HTML5 offers semantic structure, for others, only XML is sufficient.
• Modular content — chunks of content that can be assembled in different ways
• Atomic content — the smallest meaningful unit of content
• Reusable content — content that can be used multiple times in different contexts without any modification.

Consider how these terms can be used differently.  Semantically structured content does not automatically imply modularity, where the content can be reassembled in a different way.  You might use the semantic structure purely for SEO purposes, for example.  Modular chunks of content are not necessarily the smallest meaningful units, which means that the chunks may not be completely repurpose-able.  Modular content that is composed of a collection of atomic elements is not necessarily reusable content that permits the module can be used in diverse contexts.

How one thinks about these terms shapes one’s expectations for what capabilities they represent.  Some of these concepts are too new or too fluid to have a well-established meaning.  There is too much play in how they might be implemented to settle on their exact meaning and scope.  Locking down precise definitions would be counterproductive.

Structure in Linear and Nonlinear Content

The content community has expressed a range of views about the extent to which structure is compatible with narrative content.  It’s a thought provoking discussion because it touches on many core issues we grabble with.

One argument is that narrative content is different in character, and that narrative content cannot be reused.  Deane Barker has written: “To effectively manage content down to the paragraph or sentence level and re-use them in extended narratives, you would have to make sure each one was completely self-contained and match the style, tone, and tense of everything before or after. This is not easy.”  In response to this comment, Rahel Anne Bailie stated: “narrative isn’t one of the genres meant for any kinds of re-use.”

Another concern is that pre-defined structure inhibits narrative flow.  Rick Yagodich writes in his book Author Experience that “the idea of narrative and structured content may appear to be at odds with each other. Structure puts up walls and determines boxes we need to fill-in, whereas narrative is a good story, a flow that adapts to the needs of the message being conveyed.”

On the other side, some journalists are exploring how to incorporate structure into journalistic content. Adrian Holovaty, a journalist-programmer, wrote an influential post on this topic in 2006.  In it, he argued: “Newspapers need to stop the story-centric worldview.  Repurposing and aggregating information is a different story, and it requires the information to be stored atomically — and in machine-readable format.  A lot of the information that newspaper organizations collect is relentlessly structured.”  He maintained: “But stories have structure — otherwise they are a torrent of associations that aren’t logically tied together.”

There have been several journalistic initiatives to put into practice Holovaty’s ideas of making the story flow out of a structure.  The best known is Circa, which is based on the atomization of content.  Each paragraph is a distinct unit of content.  Circa differs from a lot of other storytelling through how it structures stories.  The story is emergent, and based around time, so that it builds up over time as atoms are published.

The Continuum of Structure

I question the idea that there is a clean dichotomy between narrative content, and factual (descriptive or explanatory) content.  It is true that some content, narrative especially, is predominately linear, while other content, for example an e-commerce catalog, is non-linear.   But narrative does have structure, and factual content often implies stories.

Rather than seeing two genres, story and fact, one can identify many dimensions of structure across various genres.  The below diagram shows how genres have common structures, and even different genres can have parallel structures.

What is common to many genres is a need

• to set the context of a discussion
• to establish the relevance to the reader
• to give the reader points of comparison to other things she knows about already, or might want to learn more about
• to satisfy a goal we assume the reader has
• to provide a satisfying experience, so the reader will want more content from our source in the future

Why Editorial Structure Matters to Narrators

Editorial structure — how writers and editors arrange content to provide meaning to readers — is a topic that predates intelligent content by hundreds of years.   Analog content with a strong editorial structure provides enormous intelligence for readers, even if computers can’t understand its nuances.

Semantically structured content is not the same as editorial structure. Consider a help guide. The below screen shot presents the help guide for one of the most popular XML tools intended for creating structured content.  The content is minutely structured in XML.  It can be read online, within the application itself, and as a PDF.  It would seem a prime example of the benefits of structured content.  But the content itself is barely usable.  The content is fragmented.  One cannot get a sense of the relationship among the information: there seems to be an endless list of links, some of which go to other lists of links.   When the output of structured content is so unwelcoming for audiences, many writers are understandably hesitant to embrace structured content.

The reality is that many writers and editors feel stymied by attempts to impose structure on them.  Jeff Eaton, a content developer who has worked with leading publishers, notes: “this doesn’t mean that editors and writers are content with rigid, predictable designs for the material they publish.  This challenging requirement — providing editors and writers with more control over the presentation of their content — is where many well-intentioned content models break down.”

Editors and writers concerned with presentation want more than what is offered by a CSS style sheet.  It is fundamental to their ability to communicate meaning to audiences — the deep meaning that storytelling content aims to deliver.  Dismissing these concerns as unimportant or someone else’s problem won’t advance adoption and development of structured content.  We need to appreciate and accommodate the vital function editorial structure plays.

Structure is More than Lexical

By some measures, editorial structure has become less robust with the rise of structured content.  This trend was not inevitable.  It reflects the absence of a central coordinating mechanism directing how audiences receive their content.  The separation of content from presentation and from behavior often means that none of these things is centrally coordinated.  The editor has gone missing in the action.  A core weakness stems from treating structure as entirely lexical, and assuming metadata describing words and characters are the only factor enabling structure.

Rob Waller, an information designer and fellow at London’s Royal College of Art, laments how poor the narrative experience is for a digital product compared to a printed one.   “The reader of the paper version can slip easily between related stories because cohesion within the set is provided graphically: their physical location, the typographic hierarchy, and visual genre distinctions all provide cohesion cues that in the Web version are absent or are entirely lexical.”  He notes: “whatever their actual content, we tend to assume that things that are physically close on the page are related in some way (the proximity principle), and that things that look similar are members of the same category (the similarity principle).”

Waller praises what he calls “a golden age of layout, the 1970s and 1980s. Publishers such as Time-Life, Reader’s Digest, Dorling Kindersley, and others developed a new genre that, inspired by magazine design, used the double-page spread as a unit of meaning.  The diagrammatic quality of these books – typically on hobbies, sports, history, or travel – brought layout to the fore. They were developed by multidisciplinary teams in much the same way as films are produced. Unlike the traditional book, in which the author’s voice is primary, in these books, the writer fills in spaces to order, and provides functional text such as descriptions and captions on request from editors, illustrators, photographers, and designers.”

To get a sense of how editorial structure supports narrative richness, let’s look at a couple of examples from Dorling Kindersley (DK) guides I own.  The first, from a guide to the Italian region of Umbria, presents a map of a park with associated commentary to let the reader choose their physical (or vicarious) adventure: a visit to Roman ruins, a medieval village, a summit or a cave.  The page spread shows a wide range of content: introductory explanation, the map itself, symbols on the map indicating various types of places, pictures of some of the places with a pointer to where they are, description of places with a pointer to where they are, and a sidebar of related information about wildlife in the park.  What makes the narrative rich is that it seamlessly integrates all kinds of different information types into one narrative, the story of a park.

For a very different example of editorial structure, we will look at a DK guide to the opera.  Opera is an archetypal form of storytelling, so it’s interesting to see how a narrative that’s long, complex, and multifaceted can be condensed into its essence in an engaging way.  The discussion of the opera Tosca has a wealth of structured content, but unlike much XML generated content, the structure doesn’t assault the reader.  There are information boxes with key facts about the opera performance (duration, dates of composition and first performance, librettist and sources), and the principle roles.  But the interesting structure comes from the presentation of the story itself.  Operas are structured stories, and within each act are highlights, especially the major songs.  The songs are indicated at the exact point in the story they are sung with an indication of their type (aria, duet, or ensemble), and the key line from the song in a call out.  There are also images and sidebars relating to notables performances of the opera.  Again in this example, the editorial structure leads to a planned and integrated presentation of content.

Reuse Isn’t Monolithic

What is different about the examples from the guidebooks is that the content structure seems primarily aimed at supporting audience needs, rather than reducing the burden to the publisher.  As someone who has used DK guides for many years, I am aware they use structure to reuse content across different products, and to revise their guides.  But the structure doesn’t appear to the enduser to be an efficiency measure.  Rather, it seems natural, because the elements are so well integrated.

Reuse is not the only benefit of structure.  Focusing on reuse obsessively can result in overly complicated and unworkable solutions.  We need to evaluate reuse from an editorial perspective, not just a publishing productivity one.

Content elements often have cross-dependencies.  Cross-dependencies are a good thing, even though they create challenges.  Elements offer value in relation to what they are presented with — the meaning can be based on the cross-dependency.  The integration of different elements in a thoughtful manner yields larger meaning.

The discussion of reuse can often be monolithic, looking to reuse everything, instead of selectively reusing items in the context of content that is not intended to be reusable.  When viewed from an editorial perspective, the chief benefits of reuse are to ensure accuracy when precision is essential, and to enable the combination of items in truly novel ways that bring value to audiences, rather than simply provide a minor variation.

Difference between Macrostructure and Microstructure

All structure is not the same, even when it is semantically marked up.  Some structures describe many things at once; other structures describe very specific items of information.  Discussing structure as a single abstract concept can cause us to overlook important differences.

Macrostructure is high-level structure that is common to a content type.  It provides the descriptive elements of what is being discussed.  Suppose the content deals with bird identification.  Most bird field guides have similar sections: name, identifying physical characteristics, behaviors such as feeding and nesting, habitat, voice calls, and range.

Microstructure is concerned with details and facts.  They are often the variables within a content type, and may be marked up using a standardized schema.  They identify people, places, things and quantities.

We know in many areas of life that a thing is often more than the sum of its parts — described by a scientist who pioneered the theory of the constructive emergence of hierarchies as “more is different.”  We need to understand what’s different about complex content structures.

What Bird-watching Can Teach Us About Content Structure

Birds are things in the real world that are classified with exactness.  Long before librarians thought to use taxonomies to classify content, naturalists developed the concept of taxonomies to classify birds and other living things.  One might think that birds are a topic were one can “roll-up” specific facts about a type of bird to develop larger chunks of content about them.  The challenge is that the facts about birds don’t necessarily define what they are.  They simply are indicators.  A recent book on learning (Make it Stick) notes: “To identify a bird’s family, you consider a wide range of traits like size, plumage, behavior, location, beak shape, iris color, and so on. A problem in bird identification is that members of a family share many traits in common but not all.”  It adds: “Because rules for classification can only rely on these characteristic traits rather than on defining traits (ones that hold for every member), bird classification is a matter of learning concepts and making judgments, not simply memorizing features.”

The story of a bird is more than the facts about it: it involves communicating a concept.  The difference between concepts and facts is the difference between macrostructure and microstructure.  Stories are made from both macrostructure and microstructure.

Functions of Editorial Structure

Does the content look as if it was constructed by a database?  Much structured content is not very good hiding its piecemeal origins.  And unfortunately editorial structure can’t be faked by asking your CSS expert to create a style sheet that magically makes disparate pieces of content seem like they belong together.   Editorial structure is a more comprehensive concept than font sizes and cell padding.

Where the tagging of microstructure has been motivated by search, and content reuse, the role of macrostructure is different.  Macrostructure supports how people interact with content, a major focus of editorial structure.  Editorial structure performs a curatorial role: showcasing topics (what things have in common, showing examples) and themes (comparing aspects).

Macrostructure supports way-finding.  Consider the reader’s journey.  They come from other content to this content.  What do they do next? Get more details?  Look for related content? Take action on the content? Good content has a take-away.  Editorial structure supports that.  It defines the purpose of the assembled content, while microstructure has no inherent purpose – it can be used in various situations.

The danger to narrative content is that the emphasis on smaller units of content can result in a poor narrative experience.  The issue isn’t so much whether atomic content is compatible with narrative, but how rich a narrative experience one can develop building from atomic content.

Finding the Relationships Among the Pieces

As we have become more analytical about our content, and seek to bring more transparency to the tacit judgments of editors, we can become overwhelmed by the enormity of detail we face.  Suddenly things that make sense on an intuitive level seem bewildering when exposed in minutia.

While the mechanics of intelligent content are important, it is equally important to understand how these can serve audience needs and create impact.  To do this, we need to appreciate how audiences experience content through patterns and storytelling devices.

Authors and editors use various techniques to guide the reader.  They emphasize different aspects of content.  The below chart summarizes how the choices that authors and editors make (the center two columns) can address audience needs.

On the left side are tactics from the toolkit of intelligent content.  The task is to choose appropriate tactics to support the goals of authors and editors.  Moving from left to right, each column has building blocks that support items to the right.   The building books culminate in experiences for audiences.  Conversely, starting with the needs of audiences, we can design experiences for them by building structure into content as we move toward columns on the left.

Structure is Semantic, but also Visual and Behavioral

Meaning is bigger than how something is described. It consists of implicit dimensions: perceptions and behavioral experience over time.

Perceptions are often visual, though they could be auditory, or haptic.  Visual design addresses issues like gestalt, continuation, and picture— word interactions that influence our interpretation of content.  We know from eye tracking that layout has a significant  impact on how content is perceived and understood.  It is not simply a cosmetic thing.  The term stylesheet, while a powerful concept, can falsely suggest that visual design is no more than paint-by-numbers coloration of a canvas.

Behavioral structure is a combination of interaction design (setting up how users can explore available content) and algorithmic design (the computer deciding what order and sequence of content to present).

Visuals and experiences in time are themselves information. They shape how people feel about something as much as words do.  A photo accompanying an article can dramatically shape how a reader feels about the subject. The pacing of interaction can shape how exciting or precious something seems.

We can’t allow the notion of “presentation independent” content to devolve into “experience free” content. We need to be able to describe the feelings we want our content to convey, wherever it appears.

Semantic Markup Needs Human Judgment

There is sometimes a tendency to treat semantic markup as some sort of objective reality that people uncover for the benefit of computers. This view ignores the subjective character of much semantic markup, which is essential to conveying meaning.  If semantic markup doesn’t apply judgments of humans, it is probably superficial and will be limited in what it can accomplish.

Going back to our discussion of birds.  A species doesn’t just represent a series of tagged data; it represents a concept, an idea.  There was a judgment made on how to classify the bird. Higher level structure involves judgments, which though subjective, are shared by wide numbers of people.

Like editorial structure, semantics aren’t purely lexical. People infer semantics through context and presentation.  They perceive semantic elements as having boundaries, identities, and hierarchies.  Boundaries express the aboutness of the section, which can vary in explicitness and in uniformity.  Identities may be implied, rather than explicitly named. Hierarchies must be understood to matter.

Semantic markup is not simply what is explicitly described: it is meant to capture what people interpret when they see see (or hear) the content.  The folks who understand this best are those working in the digital humanities using an XML schema called TEI.  The Text Encoding Initiative (TEI) defines markup as “any means of making explicit an interpretation of a text… it is a process of making explicit what is conjectural or implicit, a process of directing the user as to how the content of the text should be (or has been) interpreted.”  TEI uses XML to structure content to convey the meaning represented by the layout and other presentational dimensions of content. “The physical appearance of one particular printed or manuscript source may be of importance: paradoxically, one may wish to use descriptive markup to describe presentational features such as typeface, line breaks, use of whitespace and so forth.”

TEI uses metadata to describe appearance. We can similarly use metadata to express preferred presentation.

Karen McGrane has wisely counseled that “we can’t rely on visual cues” to convey semantic information. She says this because our content by necessity must be ready to be multi-channel and multi-media, and we can’t presume to know how it will appear, exactly.  But that doesn’t mean we should not use visual cues, if they are available to use. Presentation independence doesn’t mean presentation is not relevant.

Unfortunately most metadata today is exclusively literal. When used to describe legacy content, it tends to strip out meaning that is implied in the context in which it sits. When used to describe new content, it denies authors the ability to indicate the preferred context in which it might appear. We need to find ways to enhance metadata with contextual cues, so that it can convey more meaning.

While we may not be able to predict all the forms in which our content may appear, we need to think about content holistically, not just atomically.  Semantic markup should not only define boundaries, but suggest possible linkages to other semantic elements.

Making Structured Content More Narrative-Friendly

I am cautiously optimistic that semantic structure can support the development and delivery of narrative content — stories in various forms that audiences will enjoy and act on. The technical challenges are solvable. If we are to believe the view that rich narrative is the best way to gain audience attention at a time when content is too plentiful and too generic,  then monetary incentive to move in this direction is present.  But we won’t get there relying on existing approaches.  I don’t see the DITA toolkit favored by technical writers as supporting narrative content.

To enable structure to support narrative, we need to stretch our abilities in three key areas:

• Broadening our concept of narrative
• Broadening our concept of metadata structure
• Broadening our toolset

Broadening our concept of narrative

For all the interest and excitement surrounding storytelling, people often hold a surprisingly narrow view of what a story is.  For many people, a story is a plot-driven, hero-centric tale. They equate stories with the template used by Hollywood blockbusters, the hero’s journey pattern so often recycled by the advertising industry.  But stories can take many forms, and be experienced in many ways.

Stories are any content that offers a vicarious experience. The key ingredient is that people experience something: they are involved with the content.  It could be interacting with a map or a timeline, composing a plan with images, or immersing oneself in a podcast. None of these things is necessarily a story, but each of them could be. The test is asking someone what he or she did today. If they mention your content, it left a memorable impression. If they remark on some aspect that meant something to them, it indicates they experienced something using your content.

Scope is another story dimension. We tend to think about content as narrative, or non-narrative.  But it is possible to have story elements embedded in non-narrative content. One can imagine mini-stories consisting of swappable, targeted anecdotes or case studies included in a longer body of content. It’s harder to produce an experience with a short piece of content, but if appropriately targeted to be personally relevant to the audience, it could improve how audiences relate to the content.

Broadening Our Concept Of Metadata

In addition to adding element metadata, we need to expand the use of metadata describing the attributes of the elements.  If structured content is to really going to engage audiences, instead of being just more dross they have to cope with, the structure needs to reflect what is engaging about it. The success of semantically enriched content narratives will be judged and measured by the concrete impact they achieve.

Metadata needs to capture the big ideas behind the content: to indicate how we want audiences to interpret a section of content. Rather than simply indicating an “overview section,” the metadata needs to indicate what’s different about this overview section compared to others. As mentioned earlier, metadata for more complex, higher level content objects could capture more subjective, conceptual qualities, describing the fuzzier aspects of “aboutness”. Just because a quality is fuzzy (non tangible and more difficult to describe) doesn’t mean the concept isn’t real or is unimportant. When describing subjective qualities, the standard to use is intersubjective agreement: when multiple people describe a quality in similar ways (even if the exact term each uses differs). This metadata will provide valuable clues about appropriate usage of the content in different situations. I offered one idea for such an application in my CS Forum talk last year on content attractors, but there are many other applications possible.

In addition to capturing aboutness, attribute metadata should also provide clues about the compatibility of the content chunk with other chunks. Imagine you had a press release about an unfortunate event at your organization — a fire perhaps.   You note that CEO expressed his concern for the well being of people evacuated from your facility. And the press release is accompanied by a photo of the CEO — who is smiling.  Photos can have a tone, but photo metadata often doesn’t capture that. Compatibility metadata relates to any editorial aspects that indicate what items tend to work well together, or should be avoided using together. Perhaps some pull-quotes of testimonials should not be used in certain contexts: attribute metadata could indicate that.

How values are described is another aspect of metadata that can be enhanced to improve storytelling capabilities.  We tend to treat the value as a literal word that will be used everywhere. Multilingual content shows us that it is not the word describing the value that’s important, it’s what the value represents. We may call the first month of the year January, but it can be called many other things, depending on the language.  This same idea of separating values from the expression of values can be applied elsewhere.  A place can be represented by a name, geographic coordinates, or a dot on a map.  We might label an entity type on a screen using a word, an icon, or a color.  Enabling expressive fluency, where the same semantically described idea can be expressed multiple ways in different contexts, will be important to developing rich narratives using intelligent content.

Broadening Our Toolkit

Stories have structure, but that structure sometimes needs to be more flexible than used for strictly factual content to accommodate a range of expression.  Content management tools need to provide more editorial control over structure.  Cookie cutter templates make all content look the same, and dull the experience.  You can see this on Medium, where the US President’s State of the Union address looks similar to a posting by your neighbor’s college age kid.   The recent example of the Washington Post’s PageBuilder points to a possible model for giving editors more control over the structure of the content.

Narrative authors will use semantic content differently than technical communicators do. The tools need to reflect these differences.  Stories won’t be generated directly by databases. Rather, authors will use semantic content to identify appropriate content to use in different contexts. The process will change from enforced content reuse to elective reuse.  Semantic content will empower the author to find what’s available and appropriate, and create an option to include it where it can work. The goal should be to use intelligent content to empower an editor, rather than to replace an editor, to use databases to reflect curatorial decisions, rather than making those choices.

Finally, little progress will be realized until the tools for structured content become easier and more pleasant to use. That is a very big challenge given that some dimensions of content will be need to be defined in even more detail than they are today, and today’s semantic content tools are completely unacceptable to creative writers who write narrative. Again, this a solvable issue; it just hasn’t been anyone’s priority. While everyone agrees in principle that the UX of content management is ghastly, I see few people cite poor UX as a major barrier to the adoption of semantically structured content. Those who advocate using structured content for narrative, but who have largely mastered these tools already, may be unaware of, or underestimate, this chasm.

Closing Thoughts

Whether or not storytelling incorporates semantic content, and whether or not any such changes happen in the near future, the topic prompts the content strategy community to think deeply about how we approach our craft, and how it can be extended.  There is great potential, and many challenges to be solved.

— Michael Andrews