Using deep learning and word embeddings for predicting human agreeableness behavior.

Area of Science:

• Artificial Intelligence
• Computational Linguistics
• Psychology

Background:

• Deep learning advancements have revolutionized natural language processing (NLP), enabling machines to interpret and generate human language.
• The proliferation of social media has generated vast textual data, offering new avenues for analyzing human behavior and personality.
• Understanding personality traits is crucial for applications in recruitment, career counseling, and marketing.

Purpose of the Study:

• To predict the agreeableness personality trait using deep learning models.
• To differentiate between individuals who are more emotional versus those who are more logical and rational.
• To analyze behaviors associated with agreeableness, such as cooperation, friendliness, and respect for differing views.

Main Methods:

• Application of shallow machine learning, ensemble models, and state-of-the-art deep learning transformer-based models.
• Exploration of textual features including TF-IDF and POS tagging.
• Utilization of word embeddings (word2vec, GloVe) and sentence embeddings for feature engineering.

Main Results:

• The highest performance of 91.57% was achieved using sentence embeddings with the Bi-LSTM algorithm.
• This approach demonstrated superior predictive power compared to existing methods in the literature.
• Analysis confirmed the effectiveness of deep learning in decoding complex human personality traits from text.

Conclusions:

• Deep learning, particularly with sentence embeddings and Bi-LSTM, offers a powerful tool for predicting personality traits like agreeableness.
• The study highlights the potential of NLP and AI in understanding human behavior from social media data.
• Accurate personality prediction has significant implications for various fields, including human resources and consumer behavior analysis.