Linguists have long since worked out smart ways to describe accents, dialects, and characteristic vocabularies, why not use some of that aggregated knowledge to build a robust word-flow-editing system for non-experts? There’s really only one major goal of flow-editing, and that is to improve the clarity and impact of communication. The advantages that might emerge from automating flow-editing are increased understandability for non-expert flow editors and increased message-reusability and customization across different audiences.
The approach I’m currently considering involves the use of cadence and markup notes. The markup notes should provide information to the flow-editor about how to treat either certain passages or the whole document.
For example, a markup note about a paragraph could specify that the rhythm of the words should be made fast and getting stronger over time, rising to a crescendo or key point. Alternately, a markup note could specify that the author really likes a given phrase the way it is, so the flow editor shouldn’t change it at all. To go broad once more, a markup note could specify “no overly-erudite words in this whole message,” so even if you tend to slip up and get over-complex in your phrasing of a speech for 5th graders, the flow editor will do its best to keep things trim and understandable.
The cadence produced by the flow editor can be considered the system’s interpretation of the markup notes in combination with the original text, aiming at faithfully making the message “sound like” how the notes said you wanted it to sound like. To make a long story short, this should mean selecting words and phrases from a set of suitable synonyms that optimally fit the desired aural scheme.
In some ways, this is starting to sound a lot like a more verbally-focused version of LaTeX, the typesetting system that helps people write messages that are independent of the final presentation format.
