Action-Based Conversations Dataset: Task-Oriented Dialogue Systems Could Be Better
Action-Based Conversations Dataset (ABCD) is a fully-labeled dataset with over 10k human-to-human dialogues containing 55 distinct user intents. Action State Tracking (AST) keeps track of the state of the dialogue when we know that an action has taken place during that turn. Cascading Dialogue Success (CDS) measures a model’s ability to understand actions in context as a whole, which includes the context from other utterances. The novel features in ABCD challenges the industry to measure performance across two new dialogue tasks.
Making knowledge workers radically more productive with #AI for effective and efficient customer experiences.
Join Hacker Noon
Create your free account to unlock your custom reading experience.