Sentino – Methodology for Personality Profiling

This paper presents Sentino's methodology for personality assessment, which combines classical psychometric testing with semantic analysis of textual data using transformer-based NLP models. A multidimensional psychological vector space was constructed from a curated dataset of 5,000 validated psychological items spanning 20+ established inventories (Big Five, NEO PI-R, HEXACO, and others), then expanded to 500,000 items through data augmentation. A pretrained sentence-transformer model (all-MiniLM-L6-v2) was fine-tuned to project free text into this psychological space, enabling personality scoring from unstructured input such as self-descriptions, interviews, and chatbot conversations. The paper details the data pipeline, the vector-space construction, the NLP training procedure, real-world API applications (classical testing, text/interview processing, chatbot-driven assessment), and discusses limitations related to data completeness, model confidence, coreference resolution, and scope of application.