Lexical and Acoustic Deep Learning Model for Personality Recognition

Deep learning has been very successful on labeling tasks such as image classification and neural network modeling, but there has not yet been much work on using deep learning for automatic personality recognition. In this study, we propose two deep learning structures for the task of personality recognition using acoustic-prosodic, psycholinguistic and lexical features and present empirical results of several experimental configurations, including a cross-corpus condition to evaluate robustness. Our best models match or outperform state-of-the-art on the well-known myPersonality corpus and also set a new state-of-the-art performance on the more difficult CXD corpus.