Facebook Tests Surfacing Semantically Related Posts

Facebook is testing a new "mentions" feature which surfaces Pages that a user's friends were discussing, directly or indirectly.

Facebook is testing a new “mentions” feature which surfaces Pages that a user’s friends were discussing, directly or indirectly.

The feature which was first published by Oliver Chiang at Forbes, appears to be an effective way of surfacing semantically related content. The examples displayed by Oliver don’t appear to have explicit references to any given page, instead Facebook has found the most relevant page for a group of posts by him and a friend.

Mentioned Feature Screenshot

The fact that Facebook is testing the ability to determine what a status update or an article is actually about is significant. It suggests that Facebook is hard at work improving their search algorithms to surface the most relevant information whether or not that information is completely structured. While users can reference a Facebook Page by using the “@” symbol, it’s much easier if Facebook does the heavy lifting and recognizes a reference within a status or article. Below is another example presented by Oliver which demonstrates how information is not just surfaced from user updates but Page updates as well:

Mentions Page

When I wrote in April that Facebook was working on building the semantic search engine, it was based on the fact that Facebook was requiring developers to implement open graph tags within pages to present structured information to Facebook’s crawler. Now it appears that Facebook is going one step further and actually determining the text used in articles and status updates to find relevant information. Granted, Facebook has already been testing such technology, automatically tagging users as friends’ names are typed in status updates.

However this new test shows how Facebook is working hard to improve their information discovery system (the feed) by pulling together related information. Have you seen these new “mentions” feature? Do you think this is a move against Google?