Topical-Chat
- Type
- dataset
- Venue
- Amazon Alexa (github.com/alexa)
- Year
- 2026
- Source
- github
- Access
- free
- Language
- English
- Added
- 2026-07-17T20:18:03.569743+00:00
- Verified
- 2026-07-17T20:18:03.569743+00:00
Summary
Topical-Chat is a knowledge-grounded human-human conversation dataset released by Amazon Alexa, containing 235,281 utterances (over 4.7 million words) across 10,784 conversations between pairs of Amazon Mechanical Turk workers. Conversations span 8 broad topics with knowledge sourced from Wikipedia, fun facts, and news articles, with varying degrees of knowledge symmetry/asymmetry between partners. It was the largest publicly available social-conversation and knowledge dataset at release, designed to enable research in knowledge-grounded neural response generation for socialbots.
Keywords
dialogue conversational-ai knowledge-grounded open-domain socialbot nlp english mechanical-turk
Topics
NLP / Conversational AI
Research notes
- The GitHub repo is accessible and active. An enriched version with dialogue act and knowledge sentence annotations is also available (TopicalChatEnriched). Also hosted on HuggingFace as Conversational-Reasoning/Topical-Chat.