AI plus DITA/XML: a synergistic solution for knowledge base (KB) articles
During our years working together on technical and scientific projects in and around Silicon Valley, Dick and I created many structured writing solutions and training packages for software providers and educational institutions based on technologies like DITA/XML (DITA stands for Darwin Information Typing Architecture), DocBook, and Markdown.
Over the past year we have immersed ourselves in AI technology, Anna as a linguistic analyst and text annotator and Dick as an avid student of machine learning and data science.
Based on our years of experience and our newly-developed knowledge of the AI world, we have concluded that combining the traditional structured writing strategies we have used in the past with the latest AI tools and techniques has the potential to provide more powerful, cost-effective, and user-friendly KB solutions than either strategy could offer on its own.
To illustrate how these synergistic solutions could work, and the possible benefits they could provide, we have resurrected various structured documentation corpora we have written in the past, and repurposed them in a more modern, machine learning (ML)-based environment.
When we wrote our DITAinformationcenter (E8 was published in 2011) we invented for prototyping purposes two sets of short, non-technical, structured docsets called Grocery Shopping and Cleaning the Garage. We have now expanded these docsets and are transforming them into Dialogflow projects.
Dialogflow is a Palo Alto-based, Google-owned developer of human–computer interaction technologies based on machine learning and natural-language conversations. Dialogflow can be used to create conversational interfaces for websites and messaging platforms, and it is free for small prototypes, such as ours. Beyond the prototype stage, it can scale to handle larger chatbot applications.
In addition, Dialogflow has a new “Knowledge Bases” function, currently in beta, that appears to provide promise when combined with structured documentation like our DITA/XML docsets. The knowledge base functionality operates on collections of documents to create automated responses to intent requests.
Our goal in doing this project is to inspire owners and developers of technical and scientific documentation to try a new, “traditionally modern” way to significantly improve the quality and efficiency of both their knowledge base content and the delivery system that provides the information to their users.
Over the next several months, as we develop our garage, grocery shopping, and other prototypes, we will blog our experiences, tips, and techniques on this site.