Made in NY: The engineering behind social recommendations



Today, when you write a Facebook post looking for advice on local places or services and you turn on the Recommendations feature, comments to your post receive Place Cards, which highlight the suggested replies with links to relevant information, including hours of operation, phone number, address, and more. The recommended locations are also added to a map that is inserted into your original post.


We made the decision early on that we didn’t want people to have to do extra work to recommend places — no dedicated button, etc. Most of the time, you can simply post and comment, as you normally would, and this added functionality happens automatically.

In order to turn a free-text comment into a place recommendation, we have to understand the text and extract the most likely place(s) .

尝试抽出所有日记里的地点。 和网络有关的帮助中心反馈、网络诊断信息,按ASN分组。

We wanted people to be able to interact with Recommendations by commenting as they normally would, and make any necessary modifications after the comment is created, with functionality built directly into the post and comment attachments.

To solve each of these challenges, we created dynamically updating attachments.


深度学习 Meetup 总结

作者:4G Spaces 原文:

通过参加这次以及其他的一些 meetup,我总的感觉是 Google 等大鳄在深度学习方面的人才和资源积累远在其他公司之前,原因和当年大数据时代 Google 领先行业一样:Google 要解决的问题和拥有的数据远比其他小公司多。