A Novel SynergyXLBi Model to Predict Personality Trait From Text Conversation

Social media platforms serve as powerful tools for communication and promoting justice worldwide. Nevertheless, identifying trustworthy individuals and assessing strangers on social media within a short time frame can be challenging. Deriving personality traits from digital content requires mapping textual data to a personality model. The widely accepted personality assessment model is based on the Big Five traits; however, existing algorithms like Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) have limitations. To predict personality traits, we utilize bidirectional context features and extraction methods with transformer-based models. Our proposed model, Novel SynergyXLBi, combines the capabilities of Bidirectional LSTM (BiLSTM), Generalized Autoregressive Pretraining for Language Understanding (XLNet) and Conditional Random Fields (CRF). It extracts features and leverages Named Entity Recognition (NER) to classify the Five Personality Traits (OCEAN). Experimental results demonstrate that the Novel SynergyXLBi model achieves an accuracy of 97.32%, precision of 97.21%, recall of 97.24%, and an F1-score of 97.11% when compared to Random Forest, K-Nearest Neighbors, Decision Tree, and K-means classifier models. We evaluated the performance using two state-of-the-art corpora namely CoNLL-2003 and WNUT-2017, and found that the proposed Novel SynergyXLBi model surpasses the performance of existing models.