Personality Prediction using Multiple Textual Datasets and Deep Learning Models

Abstract

Personality, a core component of human behavior, determines our interactions and perceptions of the world. Recognizing and forecasting personality traits can immensely impact areas like psychology, marketing, HR, and personalized recommendation systems. Recent advancements in Natural Language Processing (NLP) have fueled a keen interest in harnessing text data, encompassing essays and social media utterances, to precisely gauge personality traits. With billions of online users generating a plethora of text data, this rich information aids in discerning personality attributes. These textual imprints, from public declarations to diverse media formats, can revolutionize our grasp of human behavior. And with millions of students going university each year fill the forms and go through the process of analytical tests. The immense strides in computing power have even enabled models to outpace human proficiency in predicting personal actions, thus having ramifications in sectors like recruitment, healthcare, and more. The allure of formulating NLP models that effortlessly decode an individual's personality traits is ever-growing. These models exploit online text, encapsulating human tendencies and inclinations, to autonomously predict personality trait levels, which holds significant real-world relevance. Consequently, the text data about a person's personality can forecast emotions based on experiences, refining systems like recommendation engines and social network analyses. This could also bolster the progress of psychological theories, leading to a more holistic understanding of human personalities. Applications in fields like marketing, human resources, recommendation systems, and social science research further underscore the immense potential of this study. The overarching aim is to harness the vast text data resources, from essays to social media posts, to get an accurate read on personality traits. This endeavor aims to enrich our comprehension of human behavior, refine decision-making frameworks, and foster the creation of smart systems that resonate with individual desires across diverse domains. Moreover, this research has achieved notable results, particularly an impressive overall accuracy of 84.4%. Detailed outcomes for specific traits include a precision of 0.97 and an F1-Score of 0.88 for cEXT using BERT, emphasizing the model's robust performance.