Reproduce “HR-MultiWOZ: A Task Oriented Dialogue (TOD) Dataset for HR LLM Agent” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Weijie Xu, Zicheng Huang, Wenxiang Hu, Xi Fang, Rajesh Kumar Cherukuri, Naumaan Nayyar, Lorenzo Malandri, Srinivasan H. Sengamedu

Researchers from Amazon and the University of Milano-Bicocca created HR-MultiWOZ, a task-oriented dialogue dataset containing 550 conversations across 10 HR domains. This dataset, generated through a pipeline combining LLMs and human input, supports the development of HR-focused conversational agents that can handle complex inquiries with empathy, addressing a critical gap in existing dialogue datasets.

Key insights

  • A multi-stage pipeline combining LLMs with human expertise can effectively generate high-quality, domain-specific dialogue datasets, particularly in sensitive areas like HR where real data is scarce.
  • HR conversations require specific characteristics such as empathy, support for long and complex entities, and domain-specific knowledge, which are not adequately addressed by general-purpose dialogue datasets.
  • Human evaluation confirms that the synthetically generated dialogues in HR-MultiWOZ are highly natural, clear, and empathetic, suitable for training robust HR-focused AI assistants.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.