Deep Learning-Based Document Modeling for Personality Detection from Text

Can language use reflect personality style? Studies examined the reliability, factor structure, and validity of written language using a word-based, computerized text analysis program. Daily diaries from 15 substance abuse inpatients, daily writing assignments from 35 students, and journal abstracts from 40 social psychologists demonstrated good internal consistency for over 36 language dimensions. Analyses of the best 15 language dimensions from essays by 838 students yielded 4 factors that replicated across written samples from another 381 students. Finally, linguistic profiles from writing samples were compared with Thematic Apperception Test coding, self-reports, and behavioral measures from 79 students and with self-reports of a 5-factor measure and health markers from more than 1,200 students. Despite modest effect sizes, the data suggest that linguistic style is an independent and meaningful way of exploring personality.