Knowledge Graph-Enabled Text-Based Automatic Personality Prediction.

Area of Science:

• Computational Linguistics
• Psychology
• Artificial Intelligence

Background:

• Understanding personality is crucial for interpersonal interactions.
• Online communication generates vast amounts of text data encoding personality traits.
• Text-based Automatic Personality Prediction (APP) aims to infer personality from written communication.

Purpose of the Study:

• To propose a novel knowledge graph-enabled approach for text-based Automatic Personality Prediction (APP).
• To enhance personality prediction accuracy by integrating external knowledge resources.
• To evaluate the effectiveness of deep learning models on embedded knowledge graph representations.

Main Methods:

• Constructing a knowledge graph from input text by matching concepts with DBpedia.
• Enriching the knowledge graph with DBpedia ontology, NRC Emotion Intensity Lexicon, and MRC psycholinguistic database.
• Embedding the enriched knowledge graph into a matrix representation.
• Utilizing deep learning models including CNN, RNN, LSTM, and BiLSTM for personality prediction.

Main Results:

• The knowledge graph-enhanced approach significantly improved prediction accuracies across all tested deep learning models.
• Enriching the text representation with external knowledge sources proved beneficial for APP.
• Deep learning architectures effectively processed the knowledge graph embeddings for personality forecasting.

Conclusions:

• The proposed knowledge graph-enabled method offers a powerful enhancement for text-based Automatic Personality Prediction.
• Integrating structured knowledge with text data improves the performance of personality prediction models.
• This approach holds promise for more accurate and nuanced understanding of individual personalities through digital communication.