Contextual Gamification Platform based Big5 Personality Trait and Preference Selection for User Recommendation

Abstract

The emergence of social networks has led to the development of many methods to evaluate people’s personalities based on their language usage and social interactions. Additionally, performance in numerous downstream tasks including sentiment analysis, text summarization, question answering, creating recommendation systems, and so forth has increased due to recent developments in deep learning-based language models. In addition to that, gamified environment attracted users more attention in any system. Gamification and the availability of social media data for personality-based preference selection have drawn the attention of researchers due to the growing need for personalized services. This research aims to examine the prediction of Big5 personality and personal preference based recommendation using text analysis, where we use a gamification based platform to understand user selection. Using a locally obtained data set, we investigate the prevalence of social network arrangement and linguistic elements concerning personality interactions. We also deploy a game based platform to assess personality based preference. We assess and compare four machine learning (ML) models and a multimodal deep learning (DL) architecture paired with Bidirectional Encoder Representations from Transformers (BERT) as a pre-trained language model as a feature extraction approach on both social media data sources and gamification platform sources. The prediction accuracy reveals that, even when evaluated with a comparable data set, the preference prediction system built using BERT-based features with Long Short-Term Memory (LSTM) and Neural Network (NN) outperforms the average baseline with a prediction accuracy of 97%. The capacity of algorithms to considerately determine a candidate’s personality-based preference to employ a game like environment and simple textual content from Facebook or textual expression opens up considerable prospects to eliminate the subjective biases associated with traditional recommendation systems.