The Relationship Between User Language and Personality Traits

Mairesse, F., & Walker, M. (2006) Words mark the nerds: Computational models of personality recognition through language. In Proceedings of the 28th Annual HLT-NAACL Annual Conference of the Cognitive Science Society (pp. 543–548). Retrieved from https://pdfs.semanticscholar.org/c3ff/cc71e87d1249b877d2b1d3f4ae4f54467ea0 .pdf

In this paper, the authors detail the findings of a study in which they examined the efficacy of using autonomous computational models to automatically assess the relationship between user language and personality traits, with the goal to develop models that can be used “to modify the output generation of a dialogue system” (p. 543). The conduct this evaluation, the authors conducted a set of experiments examining “whether automatically trained non-linear models provide better fits to the data, and whether these models can be used to recognize the personality of unseen subjects” (p. 544). The experiments collecting individual corpora and associated personality ratings, extracting relevant features from texts, and building and testing statistical models of the personality ratings based on the features on the linguistic outputs of unseen individuals. The results showed that in relation to the Big Five traits, “personality can be recognized by computers through language cues” (p. 547), with observed personality easier to model than self reports, and spoken language easier to model than written texts.  

‍