A lexical-availability-based framework from short communications for automatic personality identification. (June 2023)

• Record Type: Journal Article
• Authors: Ramírez-de-la-Rosa, Gabriela; Jiménez-Salazar, Héctor; Villatoro-Tello, Esaú; Reyes-Meza, Verónica; Rojas-Avila, Jaime
• Is Part Of: Cognitive systems research. Volume 79(2023)
• Page Start: 126
• Page End: 137
• Publication Date: 2023-06
• Subjects: Mental lexicon -- Lexical availability -- Binary text classification -- Personality identification; Cognition -- Periodicals; Cognitive engineering (System design) -- Periodicals; Artificial intelligence -- Periodicals; 153.05
• DOI: 10.1016/j.cogsys.2023.01.006
• Languages: English
• ISSNs: 1389-0417

Abstract

Abstract: We store and retrieve words from our brain during a communicative intention. These words form what linguistics and psycholinguistics refer to as the mental lexicon —a cognitive construct—. In this paper, we study the effectiveness of such a cognitive-based model for selecting the relevant lexicon that an automatic classifier can leverage for learning to distinguish personality traits from short communication intentions. Our proposed approach can automatically detect lexical units that are more suitable to train a machine learning algorithm to identify a subject's personality trait. We evaluated our method in two Mexican Spanish datasets, labeled according to the Big Five personality model. Experimental results indicate some personality traits are more transparent than others, e.g., Extroversion and Openness with 71% and 72% F-scores, respectively. An analysis of the information our proposed method selects identifies relevant psycholinguistic cues that complement psychologists' a prior knowledge. Overall, contrary to deep NNs based models, our proposed approach represents a less expensive and more interpretable technique, the desired combination for systems that aim to support the decisions made by a specialist.