Remember when context-sensitive help was the revolutionary way to deliver the right content to the right people at the right time in the right way? Just a few years ago, many technical communication teams did nothing but create context-sensitive documentation for software products. They aimed to provide contextually relevant, helpful content based on what the customer was doing in the software at any given moment.

These forward-thinking teams deconstructed large technical documents into discrete chunks, which they then hooked into the product interface. Customers no longer had to paw through a fat user manual or poke around in an online portal to seek answers to their questions. With a click of the F1 (help) key, they got the information they needed on the screen right in front of them.

Oooh. Ahhh. Contextual relevance had arrived in the digital world.

Today, savvy consumers simply expect digital content to be contextually relevant. What’s more, “context” now means more than location in a user interface. “Context” includes many factors: user-profile data, geographic location, product model, version number, preferred language, time zone, interaction history, the device’s capabilities, and so on.

Providing contextually-relevant content today is no trivial matter. It’s challenging, especially for teams that have not adopted advanced practices and tools for developing and managing information.

In his recent Content Wrangler webinar, The Fifth Element: How Structured Content Makes Chatbots Helpful, Alex Masycheff, structured-content expert and co-founder and CEO of Intuillion Ltd., discussed how emerging delivery technologies can take advantage of structured technical content to deliver contextually relevant content via conversational user interfaces, such as chatbots.

Alex delved into the following:

Read on for some highlights from Alex’s talk. For the details, go to his webinar and listen to the whole hour’s worth for free.

Single source publishing today

Single source publishing has evolved since the early days of context-sensitive help. It adapts to a range of channels. People might access it through a customer portal, through a chatbot on Facebook that provides a conversational UI, or through an augmented-reality application that applies a visual layer of information over physical objects.

Content may have to adapt also to align with business rules that determine how it gets processed. Depending on the user’s goals and preferences, access rights, and other criteria, a set of business rules can be applied, on the fly, to any content to make it deliverable to the user in a way that fits the situation.

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest
Further, we’ve broadened our notion of context sensitivity. In the early days of context-sensitive help, context meant “the user’s location in the UI.” Today, the user context has many facets. Examples:

  • Goals
  • Skills and abilities
  • Current activity
  • Profile
  • Product
  • Geographical location
  • Interaction history

Five elements of chatbot helpfulness

Alex’s webinar title starts with “The Fifth Element” in reference to the movie The Fifth Element. In that movie, four stones represent various elements in nature. A fifth stone brings them together and activates their powers.

Alex’s fifth element—structured content—brings all the others together and activates their power to create human experiences that just might qualify (depending on the human) as helpful.

Here are his five elements:

  1. User’s context
  2. User’s intent
  3. Entities of the user’s intent
  4. Knowledgebase
  5. Structured content

Element 1: User’s context

The first requisite element of chatbot helpfulness is an ability to capture info about the user’s context. The system can capture some of the contextual info (for example, the user’s location and basic profile data) automatically. The chatbot then kicks into conversation mode to “unveil” other key bits of contextual info (the user’s goal and so on).

Chatbots can gather information about people’s context by asking questions. Based on the answers they get, they can then offer advice, as shown in this conversation between a chatbot and a maintenance engineer:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

You could think of this robot as a chatty version of the old F1 key.

Element 2: User’s intent

To efficiently suss out the user’s intent—the thing someone wants to know or do in a given moment—a chatbot must keep the conversation within a narrow domain of information. Here’s an example of a domain that might support conversations between a chatbot and maintenance engineers:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

While chatbot designers can’t control what the human will toss out (ever amused yourself by messing with Siri?), they can and must define the scope of machine’s side of the conversation. Presuming that the person stays within that scope—by asking something like “Do I need to lubricate the XZ-135?”—the conversation has a chance of satisfying the user’s intent.

Element 3: Entities of the user’s intent

To understand the user’s intent, chatbots need info about the parameters, or entities, that make up the user’s intent. Here’s what such entities might look like for our maintenance conversation:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

To find out which entities go with each intent—to “fill all the required slots,” Alex says—chatbots must ask questions. For example, in the earlier conversation, after the chatbot learns that the first entity is the ZX-135, it asks a question to fill in the slot for the next entity:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

When the chatbot has filled all the entities of the user’s intent, it can proceed to offer help.

Element 4: Knowledgebase

Chatbots pull their content from a knowledge base. As content professionals, our challenge is to organize that knowledge base so that the chatbot can find and deliver the content chunks that will satisfy users’ intents.

How do we make this happen? Here’s the critical behind-the-scenes insight: Just as we have learned to structure content in standalone modules (granules), so too must we structure CONTEXT.

Aha!

Here’s how Alex illustrates a possible structure for context granules:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

Creating a chatbot is a game of matching context granules with content granules. The chatbot pulls content from the knowledge base according to that matching.

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

Element 5: Structured content

Structured content—our fifth element—unites the other elements (context, intent, entities, and knowledge base) and, as Alex put it, “activates their powers.” Structured content is granular content. In other words, it’s made up of topics (or “chunks” or “units”) that can be “managed and processed independently,” Alex says.

Without structured content, he adds, a chatbot can’t create helpful experiences.

Here’s how Alex illustrates structured content:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

To enable a chatbot to find and process the right topics at the right time, each topic must be associated with metadata that identifies applicable user contexts and user intents. Example:

  • Facebook
  • Twitter
  • LinkedIn
  • Pinterest

You might wonder why we need to bother with structure, why we can’t “just let artificial intelligence do the work.” Here’s why in Alex’s words: “We’re not there yet. Understanding human language is still a challenge.”

Watch the full webinar

For the rest of what Alex has to say on this topic—including his insights into the role of artificial intelligence, deep learning, speech recognition, image recognition, natural language processing, machine translation, metadata auto-identification, and scalability—watch the full webinar here.