Dialog System Using Real-Time Crowdsourcing and Twitter Large-Scale Corpus
来源:互联网 发布:手机淘宝如何快速秒杀 编辑:程序博客网 时间:2024/06/07 02:37
聊天系统结构如下:
(1)从web爬取料聊天语料,制作聊天对话,即Utterance Pair Corpus.
(2)用户输入聊天语句,即user input,将user input与(1)中的聊天语句匹配,当匹配置信度大于某一阈值时,得到最佳的匹配作为回答,即response.
计算与输入最相似的句子作为回答,首先计算每个句子d的词wi的向量表示:
句子d的所有词的向量组合得到句子d的向量表示.
对于两个句子d1,d2,计算其向量da,db,计算句子相似度如下:
对于输入query,首先查找本地中所有response与其的相似度,得到相似度最大的response.
(3)如步骤(2)中没有得到匹配对,则采用Real-Time Crowdsourcing,即把当前的user input 输入到其他users,让其他users来回答,得到respose给当前的usr.
阅读全文
0 0
- Dialog System Using Real-Time Crowdsourcing and Twitter Large-Scale Corpus
- 论文读书笔记-using twitter to recommend real-time topical news
- Hard and soft Real-time system
- Large-scale Incremental Processing Using Distributed Transactions and Notifications
- Real-Time SQL Monitoring using DBMS_SQLTUNE (REPORT_SQL_MONITOR, REPORT_SQL_MONITOR_LIST and REPORT_
- Real time system
- 经典论文翻译导读之《Large-scale Incremental Processing Using Distributed Transactions and Notifications》
- 经典论文翻译导读之《Large-scale Incremental Processing Using Distributed Transactions and Notifications》 转载
- Large-Scale Scene Classification Using Gist Feature
- moloch open source large scale IPv4 full PCAP capturing, indexing and database system
- Integrity Real-time Operating System
- real time cooperative editing system
- Real time H.264 Encoding and Decoding-Using FFmpeg and x264
- 翻译《real-time 3d terrain engines using C++ and DX9 》(1)
- 翻译《real-time 3d terrain engines using C++ and DX9 》(2)
- 翻译《real-time 3d terrain engines using C++ and DX9 》(3)
- 翻译《real-time 3d terrain engines using C++ and DX9 》(4)
- 翻译《real-time 3d terrain engines using C++ and DX9 》(5)
- wordBreak-ii
- 常量对象、常量成员函数及其重载
- 移动零
- Spring集成RabbitMQ-使用RabbitMQ更方便
- AsyncTaskUtilsActivity
- Dialog System Using Real-Time Crowdsourcing and Twitter Large-Scale Corpus
- WARNING: terminal is not fully functional
- hihocoder:#1520 : 古老数字
- Maven的下载和环境配置
- 计算几何之二分半平面交
- Android Studio中CodeStyle模板的配置
- 学习记录--颤抖吧,hadoop(五)---搭建完全分布式hadoop集群(1)
- 人工智能引发的四大场景革命
- JavaScript原型链与继承内容总结