Addressing Complex and Subjective Product-Related Queries with Customer Reviews-www2016-20160505

1、Information

publication：www2016

author:Julian McAuley

2、What

学习商品评论中的信息，对商品的提问，自动给出回答：按照相关程度排序

3、Dataset

Amazon 商品评论，描述，问答

4、How :

input: 1)binary question: review 的label:yes,no，eg:Is this bike a medium? My daughter is 5’8 |“This is a great bike for a tall person.” (yes)

 2)open-ended question:商品query, top-answer,review rank.  “how many hours does it keep hot and cold ?”| It doesn’t, I returned the one I purchased| “Does keep the coffee very hot for several hours.”

output:1) review是某个label的概率及label；2) review 是best answer 的概率

method：Bilinear models，MOQA

5、Evaluation:Accuray@k

详情见ppt.