In my previous post, I describe how the Tone Analyzer could make writing easier for content creators. In this post, we will look at the technology that drives the Tone Analyzer editor extension.
“The Tone Analyzer is the coolest feature in the editor. It’s helpful to students when writing an essay.”
Building the architecture
We built the Tone Analyzer feature as an extension to our TinyMCE and Textbox.io editors. First, the contents of the editor window get sent to the IBM Watson Tone Analyzer Service. IBM Watson then uses psycholinguistic analysis to process the input and returns a JSON file containing the results of the analysis. The JSON results are then processed and represented as a UI component inside the text editor.
Architecture of the tone analyzer feature
Interpreting Watson’s Results
The Tone Analyzer UI displays the results from IBM Watson as a series of emotional tones mapped to corresponding colors. The mapping between emotional tone and color was created based on existing research into color theory and emotions.
“So it is nice it explained why it has that tone, or a way to fix it if that’s the tone you wanted.”
For example, we know that a Watson score of <= 25% in the category of Social Tone: Openness is likely to mean that the reader will perceive the writing as no-nonsense and straightforward, exhibiting little subtlety or complexity. Using the Tone Analyzer feature, we represent this as “Blunt.”
From our research on color theory, we gathered that “aqua” is the most likely color to be associated with a blunt tone. We added insights into what each tone means in a tooltip that is displayed when you hover over a tone. When a tone is not present in your writing, the on-hover tooltip provides insights into how you can add more of the tone in your work.
Tone Analyzer tooltips describe the meaning of a tone when it is present (or on).
Tone Analyzer tooltips describe how to add more of a tone into your writing when the tone is not present (or off).
How Tone Analyzer stacks up in user tests
We put our Tone Analyzer to the test using the Validately service.
My next blog post looks at some of the key findings into how the Tone Analyzer can be used to make your next writing task easier.
To learn about how colors map to emotions, we researched existing works including Mapping emotion to color, Emotions through graphics depicting biometrical sensor data (Derkert et al) and The affective slider: A digital self-assessment scale for the measurement of human emotions.
This post is in a series exploring the future of content creation and the possible impacts of AI. It discusses solutions in a development or prototype state. If you’re as excited about this as we are, please reach out to firstname.lastname@example.org to discuss your interest. It will help guide our roadmap.