Scholarlink

<p>Although technological advancements enable communication through diverse mediums, conversations involving back-and-forth dialogue remain a central way in which firms interact with customers. Yet technical barriers, high subscription costs, and data security concerns hinder the extraction of insights from customer-firm conversations. To address these challenges, we create a methodological framework and end-to-end automated program, AudioMatic, designed to process audio recordings of conversations between business agents and customers. This paper details AudioMatic’s workflow, provides access to the corresponding code, and demonstrates its effectiveness using 3,900 sales prospecting calls. A logistic regression-based machine learning model trained on salespersons’ verbal and vocal features extracted via AudioMatic accurately predicts conversation outcomes with 88.08% accuracy. Follow-up analyses further reveal that salespersons’ vocal features account for 39.2% of the model’s predictive value. Overall, AudioMatic enhances the accessibility of audio analysis, highlighting both the practical and theoretical potential of integrating conversational audio data in research and practice.</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

Leveraging audio data: A guide to understanding customer-firm conversations

  • 期刊论文

  • 发表日期：

    2025年12月11日

查看全文
  • Bitty Balducci,
  • Bin Pang,
  • Lingshu Hu,
  • Can Li,
  • Wenbo Wang,
  • Yi Shang,
  • Detelina Marinova,
  • Matt Gordon

摘要

Although technological advancements enable communication through diverse mediums, conversations involving back-and-forth dialogue remain a central way in which firms interact with customers. Yet technical barriers, high subscription costs, and data security concerns hinder the extraction of insights from customer-firm conversations. To address these challenges, we create a methodological framework and end-to-end automated program, AudioMatic, designed to process audio recordings of conversations between business agents and customers. This paper details AudioMatic’s workflow, provides access to the corresponding code, and demonstrates its effectiveness using 3,900 sales prospecting calls. A logistic regression-based machine learning model trained on salespersons’ verbal and vocal features extracted via AudioMatic accurately predicts conversation outcomes with 88.08% accuracy. Follow-up analyses further reveal that salespersons’ vocal features account for 39.2% of the model’s predictive value. Overall, AudioMatic enhances the accessibility of audio analysis, highlighting both the practical and theoretical potential of integrating conversational audio data in research and practice.