fangting的博客

对话里面，有些模型是用了完整的context，有些模型是只用了上一句（我没有做完整统计），1.到底哪一种更好呢？2.如果采用第一种，能不能用好呢？Do Neural Dialog Systems Use the Conversation History Effectively? An Empirical StudySharp Nearby, Fuzzy Far Away How Neur...

1129开始

In this study, we aim to learn the meaningful distributed semantic representation, rather than focusing on building the NLU system itself. Once we obtained a reliable and reasonable semantic represen...

Because the framework can be trained in an end-to-end manner, all parameters (wi, a, b, e, f, D0, n) can be automatically learned to construct a flexible time-decay function.具体阐释universal->flexibl...

算是NLU和DST的结合。因为NLU只注意ut的slot和intent，而DST关注value的追踪（一般paper中slot是已经给定）。1）Contextual Slot Carryover for Disparate Schemas2）Improving Long Distance Slot Carryover in Spoken Dialogue Systems要解决的问题：cont...

主要针对mutlimoz2.0数据集进行学习tradesubmtcomerds-dstsom-dst流行方法：建模context成sequential形式，单个domain-slot同时作为输入----->生成或者检索，生成式：trade、comer、som-dst检索式：submt、生成检索结合：ds-dsttrade:slot-gate是指，是否需要b的gener...

CM-Net：过去的方法：1.slotfilling和intentdetect分开的，rnn或者cnn建模语句，之后用分类算法2.slotfilling和intentdetect是有关系的，Goo2018intent影响slotfilling（intent产生之后在slot生成的LSTM中加元素）zhang2018提出capsule，word->slot->intent，...