Scholarlink

<p>It can be tough to understand how a person’s words and tone influence the interviewers while conducting interviews online. Candidates are required to communicate confidently and professionally. Both verbal and nonverbal aspects of interviews are taken into account. This study proposes a system that evaluates the interviewee and scores them on three crucial factors: self-assurance, politeness, and the emotions they exhibit. The candidate’s interview video is fed as input into the framework. The framework analyzes facial characteristics from video frames and prosodic features from speech to determine a person’s level of confidence. It also utilizes the interview transcripts to assess the politeness of the language and further detects the emotions displayed using the facial features. Through the analysis report provided, the interview candidates can work on their weaknesses and improve on them. For assessing the confidence, KNN and SVR have been used with accuracies 94% and 85% respectively. The politeness score is calculated using BERT architecture and biLSTM, and the accuracy obtained is 87.1% and 89.39% respectively. Four emotions-fear, anger, happiness, and neutral-are taken into account for facial expression identification, and a CNN-LSTM is used with an accuracy of 86%. The models implemented provide scores for politeness (on a scale of 0-1) and confidence (on a scale of 0-7), and identify the emotions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Analysis of Online Interviews: A Framework for Assessing Confidence, Politeness and Emotions Portrayed

  • 期刊论文

  • 发表日期：

    2025年5月28日

查看全文
  • Aditi Nair,
  • Aditi Kamath,
  • Karishni Mehta,
  • Ruhina Karani

摘要

It can be tough to understand how a person’s words and tone influence the interviewers while conducting interviews online. Candidates are required to communicate confidently and professionally. Both verbal and nonverbal aspects of interviews are taken into account. This study proposes a system that evaluates the interviewee and scores them on three crucial factors: self-assurance, politeness, and the emotions they exhibit. The candidate’s interview video is fed as input into the framework. The framework analyzes facial characteristics from video frames and prosodic features from speech to determine a person’s level of confidence. It also utilizes the interview transcripts to assess the politeness of the language and further detects the emotions displayed using the facial features. Through the analysis report provided, the interview candidates can work on their weaknesses and improve on them. For assessing the confidence, KNN and SVR have been used with accuracies 94% and 85% respectively. The politeness score is calculated using BERT architecture and biLSTM, and the accuracy obtained is 87.1% and 89.39% respectively. Four emotions-fear, anger, happiness, and neutral-are taken into account for facial expression identification, and a CNN-LSTM is used with an accuracy of 86%. The models implemented provide scores for politeness (on a scale of 0-1) and confidence (on a scale of 0-7), and identify the emotions.