BigHeartC / Al_challenger_2018_sentiment_analysis

AI Challenger 2018 细粒度用户评论情感分析,排名17th,基于Aspect Level 思路的解决方案
327 stars 84 forks source link

LightGBM跑了20个二分类的,根据特征重要性取TopK作为Aspect #1

Closed Biedlin closed 5 years ago

Biedlin commented 5 years ago

大佬你好,请教这LightGBM跑了20个二分类的,是怎么做的呢

BigHeartC commented 5 years ago

就是对评论做分词,把词频统计作为特征,每个类别分别训练一个二分类 LightGBM 模型,判断当前评论是否属于当前类别。

Biedlin commented 5 years ago

懂了,感谢回答