Extending the Abstraction of Personality Types based on MBTI with Machine Learning and Natural Language Processing

TL;DR

This paper presents a data-centric NLP approach to predict MBTI personality types, emphasizing feature engineering and model iteration over complex models, leading to improved evaluation metrics efficiently.

Contribution

It introduces a systematic, data-focused methodology for personality type prediction that enhances interpretability and efficiency over complex deep learning models.

Findings

Attention to data quality improves evaluation metrics.

Simpler models outperform complex ones like BERT in this task.

The approach allows for broader extension of personality type abstraction.

Abstract

A data-centric approach with Natural Language Processing (NLP) to predict personality types based on the MBTI (an introspective self-assessment questionnaire that indicates different psychological preferences about how people perceive the world and make decisions) through systematic enrichment of text representation, based on the domain of the area, under the generation of features based on three types of analysis: sentimental, grammatical and aspects. The experimentation had a robust baseline of stacked models, with premature optimization of hyperparameters through grid search, with gradual feedback, for each of the four classifiers (dichotomies) of MBTI. The results showed that attention to the data iteration loop focused on quality, explanatory power and representativeness for the abstraction of more relevant/important resources for the studied phenomenon made it possible to improve the evaluation metrics results more quickly and less costly than complex models such as the LSTM or state of the art ones as BERT, as well as the importance of these results by comparisons made from various perspectives. In addition, the study demonstrated a broad spectrum for the evolution and deepening of the task and possible approaches for a greater extension of the abstraction of personality types.