Detecting Personality Traits from Texts using an Hierarchy of Tree-Transformers and Graph Attention Network with Word Embedding Refinement

Anonymous

Abstract: Automatic detection of personality traits from individuals' written texts aids professionals in evaluating mental health and individuals in identifying their strengths and weaknesses, facilitating informed decisions on personal growth, workplace compatibility, and lifestyle choices. Psychologists have discerned a collection of personality traits that can manifest within an individual's character. While BERT-based models have been successful in categorizing writings into specific personality traits, they require significant time and resources for fine-tuning. This research introduces a novel approach that utilizes a hierarchical structure of tree-transformers and a graph attention network (GAT) to classify personality traits derived from written text. It also employs an heterogeneous GAT (H-GAT) to refine Roberta word embeddings. The proposed model demonstrates substantial performance enhancements compared to previous works, as evidenced by superior results on benchmark datasets.

Paper Type: long

Research Area: NLP Applications

Contribution Types: Model analysis & interpretability

Languages Studied: English

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.