Reproduce “NLU++: A Multi-Label, Slot-Rich, Generalisable Dataset for Natural Language Understanding in Task-Oriented Dialogue” 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

Iñigo Casanueva, Ivan Vulić, Georgios P. Spithourakis, Paweł Budzianowski

NLU++ introduces a task-oriented dialogue dataset that represents user goals with reusable intent modules and slots. It covers BANKING and HOTELS, with 3,080 examples, 62 modules, and 17 slots, including shared labels. In HOTELS intent detection with about 50 training examples per fold, ROBB-QA reaches 67.4 micro-F1 versus 55.2 for ROBL-1B; this benchmark result does not establish lower cost.

Summary by alphaXiv.

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