Happy 15th birthday Google – you are now a multi-billion dollar company. Yeah I know you had your humble moments getting born in a garage somewhere in California. Imagine 15 long years! And now you greet us with your newest kid on the block. Hummingbird. So what’s it made of?
Google’s Unexpected Turn
First off, let’s greet Google’s decision to keep all the keyword search data for itself. Not a very noble move for Google. The thing is, we still get to take a peek on a chunk of that data when we’re using Adwords. The only catch is, we have to spend and Google has to earn from it. The days of Webmaster freebies celebration is coming to a close. Never thought it’ll be in September 2013.
Now I have to email all my clients telling them that the keyword search data will no longer be available. Tough stuff. How do I justify this to them? That Google had a mood swing?
In any case, I figured Bing’s still churning that data out. Perhaps it’s time to shove a little attention to Microsoft’s engine with an estimated 18% of the search market share. It’ll grow – just like how all the other backlink explorer tools grew when Yahoo shut down its backlink explorer.
So that’s that. Goodbye keyword search data!
What Makes Google’s new Engine Tick?
Just like when you buy your brand new Mustang GT and you didn’t read the manual – you’d have to still test everything out. Google’s new engine is something unfamiliar – yet it shook the search results page quite hard. The interesting thing is: we didn’t hear complaints whatsoever – at least not from the end users.
Perhaps the changes from Hummingbird is a positive one. Refining the search results to be more relevant. That is, after all, what Google really wants. A user-centric search engine.
Hummingbird, as a new search engine altogether, is made to be precise and fast. Remember Google Caffeine? Google Panda? Penguin? All those combined and refined and you get… Hummingbird. Pretty small bird for such a huge chunk of work, right?
Hummingbird is also smarter in terms of finding out the context behind the search query. It does a much better job of shuffling out your intent and giving you the best results as if you were asking your query to someone who knew all about you day in and day out. A little creepy, I know. But hey, it’s way helpful if you’re looking to get much better, relevant and comprehensive results. Every time you search, Google knows exactly what you mean and it gives you exactly what you’re looking for.
If you want to know more about Hummingbird, I find Searchengineland’s post about it really helpful.
I asked Bill Slawski of SEO by the Sea on what he thinks is the exact difference between Hummingbird and the old Google engine. Here’s his (simplified) approach on the topic:
Thanks. There’s really not a simple way to put this. It’s not a simple process. It doesn’t involve Google’s Knowledge Base results, or showing more one box results or showing more “scraped content” the way that many people are claiming it does in blog posts.
Hummingbird focuses upon a better semantic understanding of the words used within a query to reforumlate (or re-write) and expand that query, and try to get better results especially for longer queries that might be something a person speaks instead of just a number of keywords.
Google can begin this process by looking at a database of synonyms and substitute terms that it might use to replace one or more of the terms in a query with. It does this by looking at a large set of synonym or substitute rules that have been developed by looking at how different search entities might interact.
The way those synonym or substitute rules are created is through examining what people search for during query sessions, where some of the same words are used. If two queries are submitted by the same searcher consecutively, that may be a hint that the words might be considered to be synonyms or substitutes. For example, someone searches for “how to become a dentist,” and looks at some results and then searches for “how to become a dental assistant.” That means that “dentist” and “dental assistant” might be considered to be synonyms within that context.
Google might also look to see if there are a lot of co-occurring words that show up in search results for a pair of terms. Search for “car repair” and look at the top 20 results, and then search for “auto repair” and look at the top 20 results. If a lot of the same words show up in those results for both terms, they might be considered to be synonyms or substitutes for each other.
The Hummingbird process involves exploring possible candidate synonyms or substitute terms like this, but in the context of the whole query. So, if the query is something like “tell me a place to buy pizza”, Google might try to identify possible candidate replacement synonyms or substitutes for “place,” and it might then test those to see how likely it is that a synonym or substitute for “place” might fit with other meaningful terms in the query. It might ignore “tell” and “me” and “a” and “to” (calling those “skip” words) and focus upon the words “buy,” and “pizza”.
So, it might consider possible candidate synonyms such as “restaurant” or “store” or “finish” (as in someone “finishing in first place). It might combine the candidate replacement synonyms with the other words within the query, so we might see restaurant/buy, restaurant/pizza, store/buy, store/pizza, finish/buy, and finish/pizza. It might look at how often those words tend to co-occur in search results or in query suggestions, and determine that there’s a very high confidence level for “restaurant” and “pizza” to tend to appear together, above a certain threshold. Google might not see the same high level of confidence with the pairs of restaurant/buy, store/buy, store/pizza, finish/buy, and finish/pizza. So, because of the high level of confidence, Google might decide that replacing “place” with “restaurant” in the query is a good start towards answering this query. It will probably also drop the “skip” words out of the query as well. So, “tell me a place to buy pizza” is reformulated or re-written as “restaurant pizza” or “restaurant buy pizza”. Under Google’s old method of responding to a search result for a query like that, it would probably respond with local “Google Maps” results and with some local Web results as well.
So the main difference between the old approach and the new approach is that Google is using statistical language approaches to better understand how it might re-write queries to make them simpler and more likely to provide good results. It’s not looking for keyword matching as much as it would have before, so it would be concerned with only returning pages that had all of the words in the query, and would use something like “restaurant” instead of “place.
I did link in the post above to three other posts that I wrote about earlier this month and last month that discuss re-writing or reformulating queries and using co-occurrence in different ways to do that. Between this patent and those three, they point to query reformation as one of the major drivers of the kind of change that Hummingbird brings. This patent especially focuses upon announcing a query refinement or replacement like that one which was used at Google’s 15th anniversary celebration, describing Hummingbird.
So while we’re all sitting here waiting for Google to release more about the intricate details of what makes Hummingbird specifically different, let’s sit back, check our rankings and traffic, open a can of beer, and lift our glasses to tomorrow.
Here’s to a brand new Google.