[Adapt] [Seminar] How to recommend news: 2 architectures of news recommendation

龚珊三 gongshansan at sjtu.edu.cn
Wed Dec 11 14:10:08 CST 2019

Hello Adapters,

Last time I mentioned about 'session-based' news recommendation, which means user's histroy won't be saved and each session owns an unique user-id(we can also say session-id). 
This time, I want to introduce you 2 architectures of news recommendation, one is CHAMELEON, which is session-based, another is NPA, which is not. 
My aim is to help you to really understand the meaning of 'session-based' and how different it is from original item recommendation. I'll give detail explaination on tasks and architectures, no difficult background knowledge is required, hope you enjoy it.

Realated papers:
CHAMELEON: A Deep Learning Meta-Architecture for News Recommender Systems (RecSys 2018)
NPA: Neural News Recommendation with Personalized Attention(KDD 2019)

Time: Wed 4:30pm
Venue: SEIEE 3-414

See you there!

Best Regards,

More information about the Adapt mailing list