Making conversation models more empathetic

Automated dialog systems are becoming commonplace and it's important that they are able to respond to people's input in a way that is appropriate to their feelings. Recognizing feelings in the conversation partner and replying accordingly is a key communicative skill, but there hasn’t been much progress in making conversation agents more empathetic.

In our research paper, we introduce a new benchmark for empathetic dialogue generation and EmpatheticDialogues, a novel dataset of 25k conversations grounded in emotional situations. Our experiments show that dialogue models that use our dataset are perceived to be more empathetic by human evaluators, compared to previous models and datasets.

Watch the video

Read the full paper:

Towards Empathetic Open-domain Conversation Models: a New Benchmark and Dataset