Personality Prediction Based on Contextual Feature Embedding SBERT

Abstract

Personality prediction defines an individual's interior self and provides an overview of their behavioral characteristics. Individuals can develop personally and professionally with its aid. Since its inception, the MBTI has become one of the most valuable instruments available due to its widespread application in a variety of fields. Typically, psychologists use questionnaires or conduct interviews with subjects to make predictions. However, because it is only a question-and-answer session, it is prone to error. In this paper, an implicit model is suggested in order to optimize the process using machine learning. The primary objective of this paper is to use sentence transformers to discern the context of user-written social media posts in order to automate the process. In our proposed work, various text pre-processing techniques, such as data cleansing, stopword removal, and data balancing techniques such as random oversampling, are utilized. The context of the text posts is determined using Sentence-BERT (SBERT), a pre-trained model created especially for sentence-level embeddings. Using the Myers-Briggs Type Indicator (MBTI) and a variety of machine learning techniques, such as Support Vector Machines (SVM), Logistic Regression (LR), K-Nearest Neighbors (KNN) and Random Forest (RF) Classifier, it is possible to predict people's personalities based on text. SBERT combined with the Random Forest Classifier performs outstandingly to predict the MBTI personality.