A Deep Learning-based Psychometric Natural Language Processing for Credit Evaluation of Personal Characteristics
Dr.J. PraveenchandarAssistant Professor, Division of Artificial Intelligence and Machine Learning, School of Computer Science and Technology, Karunya Institute of Technology and Sciences (Deemed to be University), Coimbatore, India. praveenjpc@gmail.com0000-0002-5735-8316
S. Sankalp KarthiStudent, Department of Artificial Intelligence, Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, India. sankalp.karthi@gmail.com0009-0007-9297-576X
R. SowndharyaAssistant Professor, Department of Computer Science and Engineering, Sona College of Technology, Salem, India. rajendransowndharya@gmail.com0009-0006-6209-0357
Dr.N. Dayanand LalAssistant Professor, Department of CSE, GITAM School of Technology, GITAM (Deemed to be University), Nagadenahalli, Doddabalapura, Bengaluru, India. dnarayan@gitam.edu0000-0003-3485-9481
Debarghya BiswasAssistant Professor, Department of CS & IT, Kalinga University, Raipur, India. ku.debarghyabiswas@kalingauniversity.ac.in0009-0004-0730-9948
Manish NandyAssistant Professor, Department of CS & IT, Kalinga University, Raipur, India. ku.manishnandy@kalingauniversity.ac.in0009-0003-7578-3505
Keywords: Psychometry, Natural Language Processing, Deep Learning, Credit Evaluation.
Abstract
Psychometric assessments that gauge individuals' skills, competencies, mindsets, and personality characteristics are essential for several practical applications, including online shopping, healthcare, and cybercrime. Conventional approaches cannot collect and measure extensive psychometric characteristics promptly and discreetly. As a result, despite their significance, psychometric factors have garnered less focus from the Natural Language Processing (NLP) and data sectors. This paper presents Deep Learning (DL), the Proposed model, designed to extract psychometric variables from user-generated texts. The proposed model incorporates an innovative representation insertion, a regional insertion, a Structural Equation Modelling (SEM) encoder, and a multitasking method, all functioning collaboratively to tackle the distinct issues of obtaining complex, nuanced, and user-focused psychometric measurements. The trials on three real-world datasets involving 11 psychometric characteristics, such as confidence, nervousness, and literacy, demonstrate that Proposed model significantly surpasses conventional feature-based classifications and leading DL frameworks. Ablation research indicates that every component of the Proposed model substantially enhances its general efficacy. The findings illustrate the effectiveness of the suggested design in enabling comprehensive psychometric examination.