不得不说，F家的文章还是一如既往浓浓的工业风，这篇论文从工程角度讲解了一个召回的全流程，不管是做语义信息检索召回还是推荐召回都值得认真学习。有过前几个月的embedding召回方向工作后，深觉这篇论文对于做召回的同学来说有非常多可以总结思考的地方。

It has to be said that the article of F family is as strong as ever. This paper explains the whole process of recall from an engineering point of view. Whether it is semantic information retrieval recall or recommendation recall, it is worth learning. After working in the direction of embedding recall a few months ago, I deeply feel that this paper has a lot to summarize and think about for the students who do recall.

社交网络中的搜索除了要考虑传统web搜索中的query文本，搜索人所处的上下文也是需要重点考虑的，其中Facebook搜索场景中特有的社交图谱便是典型的一种上下文。尽管向量检索(Embedding_based Retrieval,EBR)已经广泛应用于web搜索，但是Facebook的搜索还是沿用之前的布尔匹配。这篇论文也是从升级Facebook搜索系统开始，总结Facebook将Embedding技术应用到其搜索场景过程中的踩坑经验。