[论文笔记][EMNLP-15]A Model of Zero-Shot Learning of Spoken Language Understanding

一、Model

句子和标签的对应:

the utterance ”I would like Chinese food” is labelled with inform(food=Chinese), in which inform is the dialogue action that provides the value of the attribute food that is Chinese

1.Utterance Representation Learning(输入语句编码)

2.Label Representation Learning(标签编码,标签的组成:the label aj(attk = vall) with its normal vector Waj ,attk,vall

3.Loss Function

4.Predict

For example, if the training set does not contain any examples of the act
”request(postcode)”, but many examples of ”request(phone)”, sharing the parameters can help with the recognition of ”request(postcode)” in utterances similar to ”request(phone)”

即使:比如句子”I would like Chinese food”,直接初始化其label为开篇提到的inform(food=Chinese),然后计算”I would like Chinese food”inform(food=Chinese)的点积,然后利用SVM进行多分类one-versus-rest

二、Refs

paper/others_I/others_II SVM多分类

 

0