arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2109.02020cs.CL
收藏

基于自监督学习的在线会话重入预测

Re-entry Prediction for Online Conversations via Self-Supervised Learning

The Chinese University of Hong Kong(香港中文大学);MoE Key Laboratory of High Confidence Software Technologies(教育部高可信软件技术重点实验室);Microsoft Corporation(微软公司)

Lingzhi Wang, Xingshan Zeng, Huang Hu, Kam-Fai Wong, Daxin Jiang

摘要:

近年来,在社交媒体上进行在线讨论和意见分享的全球业务正在蓬勃发展。因此,提出了再入预测任务,以帮助人们跟踪他们希望继续进行的讨论。然而,现有的工作只专注于利用聊天历史和上下文信息,而忽略了潜在的有用的学习信号,会话数据,如会话线程模式和目标用户的重复参与,这有助于更好地了解目标用户在会话中的行为。在本文中,我们提出了三个有趣的和有根据的辅助任务,即,传播模式,重复的目标用户,和转向作者,作为自我监督信号的再入预测。这些辅助任务以多任务的方式与主任务一起训练。在从Twitter和Reddit新收集的两个数据集上的实验结果表明,我们的方法优于之前的最新技术,参数更少,收敛速度更快。大量的实验和分析表明,我们提出的模型的有效性,也指出了一些关键思想,设计自我监督的任务。

英文摘要:

In recent years, world business in online discussions and opinion sharing on social media is booming. Re-entry prediction task is thus proposed to help people keep track of the discussions which they wish to continue. Nevertheless, existing works only focus on exploiting chatting history and context information, and ignore the potential useful learning signals underlying conversation data, such as conversation thread patterns and repeated engagement of target users, which help better understand the behavior of target users in conversations. In this paper, we propose three interesting and well-founded auxiliary tasks, namely, Spread Pattern, Repeated Target user, and Turn Authorship, as the self-supervised signals for re-entry prediction. These auxiliary tasks are trained together with the main task in a multi-task manner. Experimental results on two datasets newly collected from Twitter and Reddit show that our method outperforms the previous state-of-the-arts with fewer parameters and faster convergence. Extensive experiments and analysis show the effectiveness of our proposed models and also point out some key ideas in designing self-supervised tasks.

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。

↑