Scholarlink

Robotic Process Mining (RPM) automates routine discovery from UI logs for Robotic Process Automation (RPA), yet current approaches assume structured, low-noise task traces, requiring prior domain knowledge to define automatable routines. This constraint limits flexibility and applicability in real-world scenarios, as users must manually identify and record relevant tasks. We propose an approach that enables long-term user recordings without predefined routine boundaries, capturing all user interactions as they naturally occur while still identifying high-similarity, high-frequency routines suitable for RPA automation. By leveraging word2vec encoding and time-series motif discovery, our method segments task traces without prior knowledge, making the early phase of RPM more adaptive. This enhances the extraction of meaningful routines from noisy, unstructured user actions, improving automation potential in complex and evolving environments.

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

Time Series-Based Segmentation of Noisy User Interaction Logs for Robotic Process Automation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tom Hohenadl

摘要

Robotic Process Mining (RPM) automates routine discovery from UI logs for Robotic Process Automation (RPA), yet current approaches assume structured, low-noise task traces, requiring prior domain knowledge to define automatable routines. This constraint limits flexibility and applicability in real-world scenarios, as users must manually identify and record relevant tasks. We propose an approach that enables long-term user recordings without predefined routine boundaries, capturing all user interactions as they naturally occur while still identifying high-similarity, high-frequency routines suitable for RPA automation. By leveraging word2vec encoding and time-series motif discovery, our method segments task traces without prior knowledge, making the early phase of RPM more adaptive. This enhances the extraction of meaningful routines from noisy, unstructured user actions, improving automation potential in complex and evolving environments.