多任务学习在开放域对话系统中的实证研究（英文版）

英文标题：An Empirical Study of Multitask Learning to Improve Open Domain Dialogue Systems中文摘要：研究使用辅助任务对自回归模型进行改进，提高其在开放域对话生成中的长期语境处理和一致性维护，结果在小型和中型的 GPT-2 模型上表明引入新的辅助任务虽然有所改进，但仍有发展空间。英文摘要：Autoregressive models used to generate responses in open-domain dialoguesystems often struggle to take long-term context into account and to maintainconsistency over a dialogue. Previous research in open-domain dialoguegeneration has shown that the use of \emph{auxiliary tasks} can introduceinductive biases that e