Recap: SIGDOC 2020 Conference

As can be expected, many talks dealt with writing. Some talks focused around specific use cases, such as translation (NC State alumna Ashley Hardin, Aiko Sasaki, and Junko Ito of Red Hat), advanced manufacturing (David Stucker), webtext (Rob Grace and Jason Tham), and style guides (Michelle Sidler, Susan Youngblood, and Natalie Butts-Ball).

One of the more abstract talks about writing, I also oddly found the most practical. Jason Swarts’s study on the “speculative you”, besides presenting an interesting look at verbal data analytical techniques, made me consider the impact of two words that I use so often in my documentation: “if you…” Some made-up examples:

• If you have a Kubernetes cluster on classic infrastructure, you must upgrade your worker nodes before you begin.
• If you installed package 4 before you upgraded to version 3, then you must delete everything and start over.

Conditional statements like these are used more in topic-based writing than more traditional chapter-based or instructional writing. They force users to slow down (users take longer to read) and evaluate the statement, to compare it to their own experience or use case, and then decide whether the following text will be relevant to help them or not. I also think that we use a similar sort of “conditional logic” in a lot of other places, like in headings which affect navigation and tables of contents, and taxonomies for tagging content.

Because topics must stand-alone (every page is page one), conditional statements can help provide context for users to decide whether this information is helpful for them. However, keep in mind that conditional statements by nature introduce complexity, which can put your writing in contrast with Steve Krug’s exhortation to “Don’t make me think!”

Several talks presented design in different contexts that made me rethink how I view design at work. For example, Laura Gonzales, Alison Cardinal, and Emma J. Rose challenge researchers and designers to move away from an “extraction” mindset of users and their data and more towards a collaborative mindset, through users’ own language as a form of participation.

Another talk that stood out (perhaps because of COVID-19’s impact to physical workspaces) was Johndan Johnson-Eilola and Eric York’s exploration of physical offices, boundaries, and transient natures of student spaces.

But perhaps the most whimsical talk I saw (and who doesn’t need a little light-heartedness right now?) was Daniel Liddle’s “The Best-Looking Donut in Class: A Lean Approach to Visual Design in Technical Communication Courses.” He shares how a simple class assignment to have students photograph the most and least appealing pictures of donuts can get them to engage with visual considerations such as lighting, foregrounding, placement, background, and other elements that technical writers consider when illustrating (often less exciting) subject matters. It was a reminder to me that, if I find it hard to create a diagram, try to inject a little fun to get the creative juices flowing!

The video also introduced me to Lean Technical Communication, by Meredith Johnson, Michelle Simmons, and Pat Sullivan, and their 7 tenets for sustainable program innovation.

I will start out by saying that I contributed a talk about designing an analytics approach for technical content, based on IBM Design Thinking principles that my team is using to prioritize research into content analytics. The proceedings paper also has a brief lit review of about 15 articles published in STC’s Intercom on the subject, if that’s something you’re starting to look into.

However, my favorite analytics talk was a comprehensive introduction to mixed methods and tools for content analysis, presented by a panel of current and alumni NC State students Jianfen Chen, Yeqing Kong, Nupoor Ranade, and Missy F. Hannah. Their overview covers:

• Verbal data analysis for smaller sets, which is a more manual-intensive research method. An example from earlier in this article is Jason Swarts’s “speculative you” study.
• Corpus analysis for midsized and multiple sets, which is a mix of manual and computer-intensive efforts. For example, you might have a “corpus” which is all of the DITA topics for a particular product.
• Programming analysis for large and complex sets, which requires lots of programming analysis. Some common programming languages are Python and R. You might develop a script to parse your entire documentation or website content in order to detect when certain events co-occur or other sorts of analyses.
• Sentiment analysis for big data and in-flux sets, which can be hard to analyze meaning if you are new at big data analysis. You might start with a corpus, clean it up using some sort of scraping tool, and then feed it into an artificial intelligence program that does natural language understanding and other sorts of machine learning analyses.

They had three guiding tips as you begin to analyze your content:

• Methods cannot be fully determined in advance, so expect to adapt as you go.
• Careful qualitative analysis based on contextual knowledge is also important to triangulate your more quantitative findings. Quantitative findings might give you ideas for areas where to focus more and conduct more in-depth qualitative studies.
• Mixed-methods approach and triangulation within quantitative methods are beneficial, such as applying corpus and sentiment analyses.