Using Linguistic Cues for the Automatic Recognition of Personality in Conversation and Text

Abstract

It is well known that utterances convey a great deal of information about the speaker in addition to their semantic content. One such type of information consists of cues to the speaker's personality traits, the most fundamental dimension of variation between humans. Recent work explores the automatic detection of other types of pragmatic variation in text and conversation, such as emotion, deception, speaker charisma, dominance, point of view, subjectivity, opinion and sentiment. Personality affects these other aspects of linguistic production, and thus personality recognition may be useful for these tasks, in addition to many other potential applications. However, to date, there is little work on the automatic recognition of personality traits. This article reports experimental results for recognition of all Big Five personality traits, in both conversation and text, utilising both self and observer ratings of personality. While other work reports classification results, we experiment with classification, regression and ranking models. For each model, we analyse the effect of different feature sets on accuracy. Results show that for some traits, any type of statistical model performs significantly better than the baseline, but ranking models perform best overall. We also present an experiment suggesting that ranking models are more accurate than multi-class classifiers for modelling personality. In addition, recognition models trained on observed personality perform better than models trained using selfreports, and the optimal feature set depends on the personality trait. A qualitative analysis of the learned models confirms previous findings linking language and personality, while revealing many new linguistic markers.

A Survey on perceived speaker traits: Personality, likability, pathology, and the first challenge

“…Therefore, speech data should allow one to perform APP reasonably well. While still being limited, the results proposed so far in the speech literature seem to confirm the indications above for both APR (Mairesse et al, 2007;Ivanov et al, 2011) and APP (Mairesse et al, 2007;Mohammadi and Vinciarelli, 2012;Polzehl et al, 2010;Valente et al, 2012;Nass and Min Lee, 2001;Schmitz et al, 2007;Trouvain et al, 2006).…”

“…In the work of Mairesse et al (2007), APR experiments were conducted using the EAR corpus, a collection of random conversation snippets involving 96 subjects. The goal of the experiments was to predict whether each individual was in the upper or lower half of the scores observed for the Big Five traits.…”

Predicting online lecture ratings based on gesturing and vocal behavior

“…Research efforts targeted a wide spectrum of problems, including conflict detection [28], communication dynamics [7,25], mimicry measurement [10], early detection of developmental and cognitive diseases [37], role recognition [38], prediction of negotiation outcomes [9], videosurveillance [4,5,6,8], etc. Furthermore, several works were dedicated to the automatic prediction of traits likely to be relevant in a teaching context like, in particular, personality [21,23,30] and dominance [13,27,34,35].…”

“…into classes accounting for two personality traits (extraversion and locus of control). The works in [21,23] predict the way prosodic features influence the perception of personality, namely the way traits are perceived by others. Both works use machine learning algorithms (e.g., SVMs) to map basic prosodic features (e.g.…”

Social Signal Processing: The Research Agenda

“…Automatic attribution of personality traits, in terms of the "Big Five" personality model, has been attempted based on nonverbal cues such as prosody (Mairesse et al, 2007), proxemics (Zen et Given the current state of the art in automatic analysis of social signals, the focus of future research efforts in the field should be on addressing various basic research questions and on tackling the problem of context-constrained analysis of multimodal behavioural signals shown in temporal intervals. As suggested by , the latter should be treated as one complex problem rather than a number of detached problems in human sensing, context sensing, and human behaviour understanding.…”