| layout | permalink |
|---|---|
page |
/fb/ |
The FB project works on dialogue summarization and multi-turn dialogue comprehension. Previously, we have worked on contrastive fine-tuning for faithful dialogue summarization, evaluation of factual consistency, and conversation summarization benchmark. This spring semester, we will be working on multi-turn dialogue comprehension, which includes four specific tasks: dialogue summarization, response selection, conversation-based QA and conversational machine reading. We will focus on logical reasoning, knowledge grounding and commonsense reasoning to improve language models’ performance over multi-turn dialogue comprehension.