Editing Research/new digital reading experiences (section)

=== Process ===
We decided on the following workflow:<br />
First, we make editorial choices based on the source text. An example of such a decision could be changing the text design based on the primary emotion expressed with it.<br />
The next step is to set up the source text as a basic structured HTML file. Subsequently, ChatGPT is given the HTML file and asked to analyze each paragraph, analyze the content to decide on the primary emotion, and add this emotion as a class in the &lt;p&gt; tag.<br />
The last step is to use CSS to target these emotion classes and give them a specific layout.


As our first test, we asked ChatGPT to add the primary ''emotion'' fitting with each paragraph as a class in the HTML. For short passages, this worked perfectly, and its ability to use subtle cues in the text to decide the appropriate emotion is impressive. When we tried to do the whole book at once, or even chapter by chapter, it refused to annotate more than a few paragraphs at a time, being very vague about why it couldn’t process more. However, after a lot of insisting, begging and cajoling it did finally process the whole book in one go.<br />
It’s very strange to work with a tool that seems to have a will of its own. This would make it quite an unreliable tool to use in a production setting.

Our second test was to mark all ''verbs'' in the text and use CSS to style them in a specific way. Again, the annotation in HTML worked impressively. Unfortunately, it also needed a lot of persuading to actually process whole chapters at a time. We created a very basic example that can be viewed here: [https://codepen.io/arjensuijker/pen/yyJEYMN https://codepen.io/arjensuijker/pen/yyJEYMNhttps://codepen.io/arjensuijker/pen/yyJEYMN]

<span id="elective-digital-typography"></span>