Scholarlink

This chapter presents a specific application of digital trace data research in the context of conversational agents (CAs). CAs are becoming more common in industry and everyday life, creating a need for reliable evaluation methods. Building on the literature on CAs and process mining, we follow a design science research (DSR) approach. We develop a travel recommendation chatbot using an open-source framework, conduct a survey with 45 participants, and collect log data from their interactions with the chatbot. We use process mining techniques to analyze this data and conclude that conversation mining can improve the chatbot’s performance. This research demonstrates that conversation mining can supplement existing evaluation strategies, providing a more comprehensive understanding of the user experience. We suggest that this approach could be used to evaluate other chatbots and could help improve the design and implementation of future systems. Thereby, this work is a showcase example and a critical reflection of how to use digital trace data for conversation mining.

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

When Chatbots Meet Process Mining: Conversation Mining in the Era of Digital Trace Data

  • 章节

  • 发表日期：

    2026年

查看全文
  • Christian Schuler,
  • Dominik Hauser,
  • Christoph Zehendner,
  • Maren Gierlich-Joas

摘要

This chapter presents a specific application of digital trace data research in the context of conversational agents (CAs). CAs are becoming more common in industry and everyday life, creating a need for reliable evaluation methods. Building on the literature on CAs and process mining, we follow a design science research (DSR) approach. We develop a travel recommendation chatbot using an open-source framework, conduct a survey with 45 participants, and collect log data from their interactions with the chatbot. We use process mining techniques to analyze this data and conclude that conversation mining can improve the chatbot’s performance. This research demonstrates that conversation mining can supplement existing evaluation strategies, providing a more comprehensive understanding of the user experience. We suggest that this approach could be used to evaluate other chatbots and could help improve the design and implementation of future systems. Thereby, this work is a showcase example and a critical reflection of how to use digital trace data for conversation mining.