Personality Prediction System Based on Social Media Data Using Natural Language Processing

Abstract

People utilize social networking sites to express themselves, their thoughts, their professional and personal lives, their financial concerns, their food choices, the various groups and surroundings with which they engage, and many other things. These phrases can be used to characterize a person’s behavior and personality under various circumstances. We can utilize this information posted by various users across the world to determine their personality traits using the big five-factor model of personality. The big five-factor model divides personality traits into five categories i.e., openness, conscientiousness, extroversion, neuroticism, and agreeableness. The dataset used for the training of the personality prediction model is Facebook’s my personality dataset. Among various techniques used for feature extraction from texts, we got maximum accuracy by using TF-IDF with SVM. Although we applied different combinations of feature extraction methods with different classifiers like using hashing, bag of words, BERT, glove for feature extraction with Multi-Layer Perceptron, Decision Tree, Random Forest, Logistic Regression, and Gaussian Naïve Bayes, etc., after training our prediction model we got average accuracy of five traits 69 percent. Now to utilize this personality prediction model we integrated it with a traditional recruitment system to make it a smart one. We integrated this model into a system called “Smart Recruiter”. It is a web-based application that provides services to both employers and job seekers both. Job seekers can use it to search for jobs offered by different employers and employers can use it to post different jobs and search for candidates for themselves, but some services for employers are paid and they need to subscribe to it by making payment via the integrated payment method i.e., PayPal. But what makes smart recruiter different smart from other traditional recruitment systems is that it utilizes the job seekers’ personality traits which are predicted using tweets posted on their profile and then the traits of the users are matched with the traits required for the job post and the user is shortlisted automatically if they are matched. Moreover, we to fetch real-time tweets of job seekers and authenticate the username given by job seeker if exists on Twitter or not, we have used Twitter API for that purpose