LEDA: a Large-Organization Email-Based Decision-Dialogue-Act Analysis Dataset

Mladen Karan, Prashant Khare, Ravi Shekhar, Stephen McQuistin, Ignacio Castro, Gareth Tyson, Colin Perkins, Patrick G.T. Healey, Matthew Purver

Open on GitHub

Type: dataset
Venue: ACL 2023 (Findings) / sodestream project
Year: 2023
Source: github
Access: free
Language: English
Added: 2026-07-17T20:18:03.576070+00:00
Verified: 2026-07-17T20:18:03.576070+00:00

Summary

LEDA (Large-organization Email-based Decision-dialogue-act Analysis dataset) is a dataset of email conversations from the public mail archives of the Internet Engineering Task Force (IETF), annotated with dialogue act labels to study decision-making in large distributed organizations. Emails are split into segments and labeled by trained linguist annotators with a multi-label taxonomy of dialogue acts. The dataset includes a BERT-based baseline model for dialogue act tagging. Published in Findings of ACL 2023, it addresses the gap between face-to-face meeting datasets (AMI/ICSI) and asynchronous online collaborative communication at scale.

Keywords

email dialogue-acts ietf decision-making nlp annotation bert acl2023

Topics

NLP / Dialogue Acts / Email Analysis

Research notes

GitHub repo is accessible. The project is part of the 'Sodestream: Streamlining Social Decision Making for Enhanced Internet Standards' EPSRC project. Authors are from Queen Mary University of London, University of Essex, University of Glasgow, and HKUST.