Through the convergence of disability studies and literary cognitive studies, Autistic Characters: (de)coding embedded sentiment explores depictions of autistic characters in literature with the use of close readings and scaled readings, a computational analytics method which uses sentiment analysis to decode the sentiment embedded in texts. I investigate these characters through close readings in which I explore my positionality within the major fields of study and the embedded medical and social histories coded into neuroatypical and neurodiverse literary representations of autism. Building upon the perspectives of my positionality and these histories, I explore how the substrate of literature is coded for a neurotypical and ableist focused reading. In my continued exploration of the embedded sentiment in literary constructions, I build upon the traditional close readings of autistic characters as I expand this analysis to conduct a (de)coding by scaled readings through which I produce visual representations from net sentiment (positive minus negative), total sentiment (positive plus absolute value of negative), negative sentiment, and positive sentiment measurements. These sets of visualizations are created both by chapters and in evenly spaced 500-word intervals throughout a full-length novel. To generate these scaled readings through the digital humanities method of sentiment analysis with the lexicon "bing," I use the programming language "R" to reveal the sentiment that lies latent within the texts. The visual patterns that emerge from the scaled readings provide graphical depictions from the positive and negative sentiment which allows me to re-read the text to analyze how it is coded with patterns, providing both a precise and different reading. I then further explore the origins of the code in the sentiment lexicon "bing" that generates the "positive" and "negative" data points. In this exploration, I critically examine the accuracy of this method and problematic constructions that arise from human generated lists that are used by machine learning to gauge the sentiment of words. Yet despite inaccuracies that may arise with scaled readings in combination with the biases of the lexicons, the visual patterns provide for a method of re-reading with sentiment that has not yet been explored. A method of reading that can lead to a different understanding of how the positive and negative embedded substrate generates charged sentiments which contribute to priming narrative feelings and in turn influences receptions of autistic characters.
Read
