Towards Empathetic Open-domain Conversation Models: a New Benchmark and Dataset
This work addresses the problem of building more empathetic AI conversation systems for users, though it is incremental as it builds on existing dialogue models and datasets.
The authors tackled the challenge of creating AI dialogue agents that recognize and respond to human emotions by introducing a new benchmark and the EmpatheticDialogues dataset of 25k conversations. Their experiments showed that models trained on this dataset were perceived as more empathetic by human evaluators compared to those using general internet data.
One challenge for dialogue agents is recognizing feelings in the conversation partner and replying accordingly, a key communicative skill. While it is straightforward for humans to recognize and acknowledge others' feelings in a conversation, this is a significant challenge for AI systems due to the paucity of suitable publicly-available datasets for training and evaluation. This work proposes a new benchmark for empathetic dialogue generation and EmpatheticDialogues, a novel dataset of 25k conversations grounded in emotional situations. Our experiments indicate that dialogue models that use our dataset are perceived to be more empathetic by human evaluators, compared to models merely trained on large-scale Internet conversation data. We also present empirical comparisons of dialogue model adaptations for empathetic responding, leveraging existing models or datasets without requiring lengthy re-training of the full model.