Computational Inference of Candidates' Oratory Performance in Employment Interviews Based on Candidates' Vocal Analysis

Employment interviews is a popular tool to select appropriate job candidates for the open position. Understanding the important factors that are assessed in the employment interviews has become important in organizational psychology. We propose a system which will assess how well did the candidate speak during the interview based on the candidates' voice. This paper presents a computational framework for the automatic prediction of the oratory performance of the candidate through vocal analysis. We have applied multi-classification algorithms on an audio dataset where candidates are subjected to five psychometric questionnaires during job interviews. Our framework automatically predicts the ratings for constructs like Confidence, Fluency, Energetic and Clarity with an accuracy of 70 % and higher, that constitutes to be the major four features of our framework. We then predict the overall oratory performance of the candidate thereby defining a measurement term for oratory analysis called as Oration Index. In the last step, we analyze and find the important vocal attributes that contribute in prediction of Oration Index constructs. Thus, our proposed framework will evaluate the speaking performance of the interviewee based on his/her voice using multi-classification algorithms.