The back and forth between champions of DITA and Markdown is nothing new. Both are useful, but DITA’s structure is superior for maintaining large knowledge bases ready for complex systems like chatbots.
Recently, we were asked about metadata in Markdown compared to metadata in DITA XML related to deploying knowledge content to a chatbot. Metadata is information about information, and it’s vital for organizing large content repositories. It’s also a major component in deploying to systems with high semantic requirements, such as chatbots.
Put plainly, you want to be able to do two things with your metadata:
- Find the content you’re looking for in your library
- Deliver accurate information from your content library to the end-user
It sounds simple, though it’s anything but. This is especially true as technologies for information delivery continuously evolve. As industry 4.0 continues to bring new technologies and ways to search, synthesize, deliver information, where information is stored and how it’s subsequently parsed and matched to a user’s query requires structure.
The topic, as a whole, is vast, which is why we’re going to narrow our focus on deploying content to chatbots, then to the end-user, and how your content structure is crucial to the success or failure of that process.
Chatbot 101: A Few Things to Get You Started
There are some points that need to be made about chatbots before diving into why DITA has a more apt metadata structure for them. Chances are you’ve interacted with several chatbots while using your smartphone or computer, but understanding what they are, what they’re not, how they work, and what they’re capable and incapable of will foster a better understanding of why good metadata is so important to them functioning properly.
What is a chatbot?
Chatbots are conversational interfaces that provide information in human-like responses in real-time. Built around content delivery, chatbots are built to find answers to questions, solve problems, deliver content, and otherwise provide an interactive way to engage with customers.
Of course, chatbots can serve diverse functions that you’ll need to define according to your business needs. For our purposes, chatbots are functional chat-based applications that are meant to improve customer experience by mirroring how humans converse with one another via messaging platforms.
What isn’t a chatbot?
They’re not replacements for human beings. They should provide the best information possible in their interactions, but know when to defer to human beings when a query is outside their scope or knowledge base.
It’s not wise to attempt to build a chatbot disguised as a human being. Your end-users are smarter than that and will be able to break through that facade very quickly and, chances are, they won’t be thrilled about it. Chatbots and human beings are meant to work together to make a user’s experience better.
Why the rise in chatbot popularity?
They’re popular because they streamline aspects of customer experience that were formerly handled by human beings. With chatbots able to interact with customers for certain tasks, human personnel can focus their efforts on other meaningful work.
It’s also worth mentioning that billions of people text and use messaging apps on a daily basis. As a society, this is one of our most comfortable and preferred methods of communication. Why wouldn’t a business take advantage of this? Chatbots make conversational text-like interactions an easy go-to that people are used to and comfortable with.
How does a chatbot work?
At the most basic level, and for our purposes now, chatbots have a framework structured like this:
- User Interface (UI): The user interface is what customers on your website see. That chatbot bubble that they can type a query directly into. Like the one at the bottom right of our home page.
- Chatbot Engine: Chatbot engines take the text a user types and match it to predefined user intent.
- Knowledge Repository: This is your content library, the place the chatbot looks for answers to a user query.
In essence, the system looks like something like this:
The problem we face in making our chatbots useful for customers and valuable to us is ensuring they can find what they need in a large content repository. Easier said than done, it’s semantically rich metadata that helps chatbots successfully parse vast content libraries to find solutions that match user intent and answer their query. That’s where DITA outplays Markdown by leaps and bounds.
Metadata: The TL;DR Your Chatbot Understands
Too long; didn’t read. At the top of long articles, sometimes web writers will put a sentence or two that gathers the gist, context, and main idea of the article. That way, a reader can decide if the content is relevant to what they’re looking for without reading the whole thing to find out.
Metadata works in a similar fashion, but for machines. Bad metadata is a poorly defined tl;dr that leaves a machine reader unsure of what the content is about. Naturally, neither machines nor humans derive much use from this. Well-defined, semantically meaningful metadata is an effective tl;dr that defines what the content is and if it’s applicable to a user query.
Markdown and DITA both support metadata, but their respective methods of implementation and overall capabilities, are starkly different. With complex systems like chatbots, metadata needs to be structurally sound and semantically meaningful.
For your chatbot to know things, it needs to be able to find things. Without well-defined semantic metadata, you’re not doing it any favors. This will result in a frustrating chatbot, which directly correlates with a frustrating user experience. Let’s avoid that.
Markdown: Short Term Good, Long Term Bad
Metadata in Markdown syntax is relatively simple, but it remains largely applied at the document level. This is fine for smaller scale documents and projects intended for singular use, though it’s difficult to maintain as documentation libraries expand. So, yes, metadata in Markdown is useful and applicable, but it tends to fall apart around structural enforcement.
This is problematic once there are numerous authors across a sweeping library of content. Without enforcing and ensuring Markdown metadata conventional standards are applied to your internal content development processes, authors are largely left to their own devices. This can be disastrous when your chatbot is parsing through a content library with inconsistent or poorly defined metadata.
DITA: Structured with Semantics in Mind
When we talk about semantics, we’re really talking about how we create content meaning that machines will easily understand. Computers don’t work like human beings, so defining meaning in a way machines can synthesize requires structure and rich semantics. Chatbots are no exception, especially as your knowledge repository grows. The bigger the knowledge base, the more deliberate, semantically rich, and organized your metadata needs to be to ensure your chatbot is able to find and deliver relevant information.
Where DITA shines is that it’s an accepted XML standard with a documented infrastructure. Markdown metadata lacks that structural standardization. Because of standardization and inherent structure, DITA is a more practical option for businesses that are posturing their content libraries to be deployed by chatbots. DITA is built on a foundation poised for scalability. DITA future proofs your content by making sure the semantic metadata is rich and structural conventions are adhered to.
Why Chatbot Competence Matters
It’s no secret that we exist in a mobile-first digital landscape and among those billions of mobile users, messaging apps remain far-and-away the most popular. At this point, not having a conversational experience available to your customers through a chatbot is neglecting prospects in one of the most common interactive modes across the planet.
However, we have to reiterate the importance of the knowledge repository and information structure behind your chatbot. Your chatbot can only be as good as the foundational knowledge you’ve given it to work from. That’s why high semantic metadata is important for your bot’s ability to find and deliver accurate information from your content repository.
Chatbots are fascinating, but they’re not magic. The work you put into organizing the information behind them will define their abilities and, ultimately, shape how your customers experience them.
At the end of the day, you want whatever you build -- foundation, structure, and process -- to support your product’s eventual size, not the minimum viable product. When it comes to substantial content repositories aiming for content delivery through chatbots, DITA remains the most suitable choice.
