Texture is an experimental program that generatively designs posters based on linguistic sentiment
What gives language its sentiment?
In his book The Secret Life of Pronouns, James W. Pennebaker explores sentiment analysis in-depth. While we often associate tone and sentiment with action words and key nouns, Pennebaker concluded that auxiliary words and pronouns actually have a significant impact on the tone and style of language. The usage and commonality of different types of auxiliary words used in a text dictate how logical or emotional the writing comes across. The use of pronouns can also indicate the writer’s social status and emotional state.
I looked at how other programs analyze sentiment with the use of word affinity resources. Combined with other linguistics research, I developed a few key metrics of analyzing text.
- Positivity & Negativity
- Sociability & Intimacy
- Orderliness & Chaos
- Distinctiveness & Commonality
How can computers evaluate emotions in writing?
The making of Texture can be broken down into two parts. First, I wrote literary analysis code that detects the genre, tone, sentiment, and status of any given text by analyzing pronouns abd auxiliary words . Those results are each mapped onto a sentiment scale between -10 and 10.
The sentiment scale contains four major sets. A given text will have a value in between the two sentiment extremes.
In use example:
From literary analysis to generative design
I created visual designs based on my literary sentiment values. The design of the posters is purely generative. Each sentiment value is directly and indirectly correlated with composition, scale, shape, color, and contrast.