End to end Memory Networks with Knowledge Carryover for Multi-Turn Sporken Language Understanding
2020-02-08 02:28
477 查看
摘要
这是一篇讲述在利用端到端记忆网络做多轮次的口语语言理解的论文。里面提及利用多轮次的信息主要有两个挑战：1、上X轮的错误有可能会传送到后面的轮次。2、上X论的信息可能会忘记。本文采用对对话和对应的意图以及槽信息都编码为记忆，然后用注意力机制进行解码，利用储存的语义同时应用在意图分类和槽信息填充的任务中。在微软小娜的数据集上比单轮的rnn有更加的效果。
小评
用编码的方式实现对多轮次的对话信息的保存是一个今年比较常用的方法，而且这种方法的网络结构相对简单，符合奥卡姆剃刀原理。但是使用的是非公开数据集。有不同的结构的比较但是没有将研究进展重现在私有数据集中不得不说有点遗憾。我还是比较想看到他这种模型和Yang的双向LSTM的网络进行对比，两者都是对多伦茨的对话进行了处理，Yang的网络跑起来太费劲了。
- 点赞
- 收藏
- 分享
- 文章举报
seeledu
发布了3 篇原创文章 · 获赞 0 · 访问量 241
私信
关注
内容来自用户分享和网络整理，不保证内容的准确性，如有侵权内容，可联系管理员处理 
相关文章推荐
- 【论文笔记】An End-to-End Model for QA over KBs with Cross-Attention Combining Global Knowledge
- Reasoning with Memory Augmented Neural Networks for Language Comprehension
- An End-to-End System for Unconstrained Face Verification with Deep Convolutional Neural Networks
- 1606.End--to-End Comparative Attention Networks for Person Re-identification 论文笔记
- 神经网络 | DeepVO:Towards End-to-End Visual Odometry with Deep Recurrent Convolutional Neural Networks框架
- Reading table information for completion of table and column names You can turn off this feature to get a quicker startup with -A
- End-To-End Memory Networks 论文阅读
- Tao Yu的论文之一 "Service Selection Algorithms for Web Services with End-to-end QoS Constraints (ISeBM2005)"
- End-To-End Memory Networks
- 论文：记忆网络之Gated End-to-End Memory Networks
- 读论文end-to-end Memory Networks
- End-To-End Memory Networks 论文翻译
- 论文阅读：End-To-End Memory Networks
- How to build Multi-Language Web Sites with ASP.NET 2.0 and VS.Net 2005
- How to build Multi-Language Web Sites with ASP.NET 2.0 and VS.Net 2005
- [论文笔记] SOA Middleware Support for Service Process Reconfiguration with End-to-End QoS Constraints (ICWS 2009)
- Convolutional Patch Networks with Spatial Prior for Road Detection and Urban Scene Understanding
- 车牌检测识别--Towards End-to-End Car License Plates Detection and Recognition with Deep Neural Networks
- QA： Dynamic Memory Networks for Natural Language Processing
- Multi-tasking RTOS for microprocessors with limited memory by saving only a single return address per task during context switching