Google Hummingbird - the Beginning of Latent Search

Most of us have had a chance to experiment with Siri on an iPhone or some of the Voice Assistants on an Android mobile device. You may not have experimented with something Google rolled over the past few weeks code named Humingbird though. It's a major new core algorithm for Google that allows users to use conversational speech for searching. I'm amazed at how under the radar this has been. Different terms are being tossed around for this including "latent", "conversational" and "abstract" - here's a quick 2 minute video I recorded demonstrating how it works.


We're so used to searching on keywords it's second nature. Larry Kim from Wordstream refers to keyword searches as "Caveman English".  Hummingbird is probably the biggest change in Google's search technology since 2001. It's similar to Facebook's Graph Search, allowing users to use more abstract or latent language when searching - the same kinds of things we do when having a conversation with another person.

It's also something Google has to do to stay competitive. Both Facebook and Siri use Microsoft's Bing for searching with Apple just switching from Google to Bing with iOS7. Current estimates put Google's search market share at around 70% with many referring to the company as a search business that also does experiments.

If Apple gets Siri fixed up and Facebook also fixes up Graph Search, Google could pretty rapidly lose search market share. Facebook has to be a huge concern right now - recent comScore reports that analyze Americans’ surfing patterns found people are spending more time on Facebook than Google. If Facebook can get their search act together (it's pretty bad right now) lookout.

Right now, Google's Voice Search sure feels pretty natural to me - grab the latest version of the Chrome browser and give it a try.