A sentiment-aware deep learning approach for personality detection from text

Personality detection based on user-generated text content analysis has a significant impact on information science, for instance, information seeking. Existing deep learning-based approaches, however, have two major limitations. Firstly, they extract only keywords for personality detection and lack the analysis of sentiment information and psycholinguistic features. Secondly, the information about the context and polysemous words are ignored. To tackle these problems, we propose a novel multi-label personality detection model based on neural networks, which combines emotional and semantic features. Specifically, we leverage Bidirectional Encoder Representation from Transformers (BERT) to generate sentence-level embedding for text semantic extraction. In addition, a sentiment dictionary is used for text sentiment analysis in order to consider sentiment information. Finally, we input the above semantic information and emotional information into the neural network to construct an automatic personality detection model. The performance of the model has been evaluated on two public personality datasets. The experiments show that we obtain average accuracy improvements of 6.91% and 6.04% on the Myers-Briggs Type Indicator (MBTI) and Big Five datasets, respectively, compared with the state-of-the-art techniques.