《论文阅读》使用条件变分自动编码器学习神经对话模型的语篇水平多样性 2017 ACL 前言简介相关知识Stochastic Gradient Variational BayesMultivariate Gaussian DistributionIsotropic Gaussian DistributionReparameterization Trickprior network & post

1. 选出最佳答案 1.On the contrary 2.Despite that 3.Mind you 4.Look 5.On the contrary 6.After all 7.Still 8.Mind you

1. 选词填空 1.granted, but 2.certainly, still 3. 3.Of course, Even so 4.it’s true, Nonetheless 5.It is true, Nonetheless

论文地址：https://www.aclweb.org/anthology/2020.acl-main.551.pdf 本文的起点 最近的抽象式摘要都是对于提前抽取的每个句子进行精简或者重写，但是一般来讲，有些句子是连贯的，例如需要合并2个句子为1个句子。 想去做一个新的摘要方法，相比较句子级别的摘要，它能够更有信息量，也更精简。 待解决的问题，一个是哪些EDU应该被挑选出来；另一个问题是

1. 圈出最佳答案 1.First of all 2.Regarding 3.Now 4.Right 5.In conclusion 6.By the way 7.For no thing 8.As far as the repairs are concerned. 9.Talking about, finally, In short

1. 填空训练 1.Well 2.as a matter of fact 3.In fact / Well 4.actuallly 5.In fact 6.In fact

1. 划掉错误答案 1.Furthermore/Moreover 2.In particular / For example 3.That is to say / I mean 4.besides / another is 5.Indeed / That is to say 6.In particular

1. 选词填空 1.frankly 2.I’m friaid 3.At least 4.I suppose 5.Let me see 6.Honestly 7.No doubt 8.Apparently 9.I’m afraid 10.or rather

https://blog.csdn.net/sinat_26917383/article/details/55683843#commentBox