arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1902.00098cs.AIcs.CLcs.HC
收藏

第二届对话智能挑战赛(ConvAI2)

The Second Conversational Intelligence Challenge (ConvAI2)

Facebook AI Research(Facebook人工智能研究);Moscow Institute of Physics and Technology(莫斯科物理技术学院);University of Montreal(蒙特利尔大学);McGill University(麦吉尔大学);Carnegie Mellon University(卡内基梅隆大学);Microsoft Research(微软研究院)
Emily Dinan, Varvara Logacheva, Valentin Malykh, Alexander Miller, Kurt Shuster, Jack Urbanek, Douwe Kiela, Arthur Szlam, Iulian Serban, Ryan Lowe, Shrimai Prab… 展开作者

Emily Dinan, Varvara Logacheva, Valentin Malykh, Alexander Miller, Kurt Shuster, Jack Urbanek, Douwe Kiela, Arthur Szlam, Iulian Serban, Ryan Lowe, Shrimai Prabhumoye, Alan W Black, Alexander Rudnicky, Jason Williams, Joelle Pineau, Mikhail Burtsev, Jason Weston

摘要:

我们描述了ConvAI 2 NeurIPS竞赛的设置和结果,该竞赛旨在进一步发展开放域聊天机器人的最新技术。比赛的一些关键要点是:(i)预训练的Transformer变体是目前在该任务上表现最好的模型,(ii)但是为了提高与人类的多轮对话的性能,未来的系统必须超越诸如困惑度的单个单词度量来测量跨话语序列的性能(对话)--在重复、一致性和对话行为的平衡方面(例如,问了多少问题,回答了多少问题)。

英文摘要:

We describe the setting and results of the ConvAI2 NeurIPS competition that aims to further the state-of-the-art in open-domain chatbots. Some key takeaways from the competition are: (i) pretrained Transformer variants are currently the best performing models on this task, (ii) but to improve performance on multi-turn conversations with humans, future systems must go beyond single word metrics like perplexity to measure the performance across sequences of utterances (conversations) -- in terms of repetition, consistency and balance of dialogue acts (e.g. how many questions asked vs. answered).

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑