Wednesday, April 23, 2014
   
Text Size
  • Home
    Home This is where you can find all the blog posts throughout the site.
  • Categories
    Categories Displays a list of categories from this blog.
  • Tags
    Tags Displays a list of tags that has been used in the blog.
  • Bloggers
    Bloggers Search for your favorite blogger from this site.
  • Team Blogs
    Team Blogs Find your favorite team blogs here.
  • Login

A Search Geek's Thoughts on Google Hummingbird

Posted by on in Industry News
  • Font size: Larger Smaller
  • Hits: 14496
  • 9 Comments
  • Subscribe to this entry
  • Print
  • PDF

Yea, I'm going there. Normally I don't jump on the bang wagon that is the flailing of the SEO industry each time Google rolls over in the night. The reasoning lies in that we generally don't know much early on and it takes time to digest things. Today, I am breaking that rule.

While I don't really have any hard insights into Google's new Hummingbird algorithm change, we have been talking amongst ourselves about some things over the last month, that are now making some more sense.

Google-HummingBird.jpg

 

Changing how they deal with queries

A while back Terry and I were looking at some query spaces we were intimate with and noticed that something was indeed changing in how Google was doing their query classification. Of note, there seemed to be a different mix of transactional and informational results in spaces that had previously been transactional heavy. Something was afoot...

From what we do know so far, one of the elements of the Hummingbird is trying to deal with the myriad of ways that a user might go about a search task, (chain of queries to accomplish a goal).

Over the last 10 years we've seen how shorter queries have given way to longer ones. I have on occasion take my laptop to the local pub and asked people to find something (on Google). Everyone goes about it in different ways as far as the queries they use to get there. We also see that people have evolved with search and tend to use more words and be far more specific in how they query Google.

This can be a problem for a search engine that relies on keywords. Some of the elements we are seeing with the latest update do indeed seem to be dealing with query analysis. I have been talking about it in terms of 'synonymous queries'. Meaning; that while the words used in a query might differ, the goal is the same. Google might be looking at treating different queries as being the same as far as the goal of the task.

Consider queries like;

  • [where to get blue widgets]
  • [where can I find blue widgets]
  • [where to buy blue widgets]
  • [where can I buy blue widgets]


You get the idea. In the past, the subtle differences in the wording might actually produce different results. From what I am seeing so far, treating these kinds of queries the same, or as being more closely related, could be part of what has changed with Hummingbird.

 

The Conversational Search Connection

This I think is one of the more important elements as far as where Google seems to be headed. Obviously the foray into Android and eventual purchase of Motorola was the writing on the wall as far as Google's interest in the world of mobile.

The obvious evolution there of course is voice search. Who wants to sit there typing queries into Google on that small screen? I know I don't. So it makes sense. Now, let's also consider the recent work they did with conversational search, including voice abilities. With this they really highlight how connecting natural language and search tasks are being taken to a new level.

As an example, you can try (via Chrome);

  • [how old is Barack Obama?] - which returns a knowledge card with age and birthday
  • [who is his wife?] - returns knowledge card for Michelle Obama
  • [where was she born?] - returns a knowledge card for Chicago


Conversational-1.jpg

Conversational-2.jpg

b2ap3_thumbnail_Conversational-3.jpg

Originally this was just pulling from the knowledge graph bits, but is now apparently available for the entire search core.

With this we can see how the previous search was extended into the new queries in the search task chain. And the obvious implications of natural language processing in the voice elements. People simply don't speak in keywords. Which will be key in dealing with voice/mobile search in the years to come.

As noted in this Forbes article;

“After the event, Scott Huffman, a key engineering director at Google currently working on natural language, told me that part of the impetus for the change was that as more people speak searches into phones, they’re doing so in a more natural way than they type in queries–which is to say more complicated. So Google’s search formulas needed to be able to respond to them. “

 

The Nuts and Bolts; Things not strings

If there's one thing we've been talking about here in the Dojo for the last while.... it's about keywords. Or, more succinctly, the end of them.

Over the years search has become less and less about keywords and more about concepts. The days of keyword stuffing are long gone. Using synonyms, is not what 'semantic search' is all about. Modern search engines seek to try and understand the concepts and relations of the content on a page. This post from 2010, shows how long I've been rambling about it...sigh...

Anyway, don't wanna pull a muscle trying to pat myself on the back, so let's move along shall we?

It's not a huge leap for us to start to see this manifesting more with Hummingbird. Through a combination of natural language processing, query data and semantic analysis, there is a far more implicit behavioural element that is becoming part of the natural evolution.

Bill had a great post the other day about one patent that might be part of this evolution. It deals with some elements (via query analysis) that can surface content that satisfies the user, although they maybe used different approaches (queries and query reformatting) to reach the same end result. The old Google would, for the most part, see the words in the query, not really trying to establish the implicit elements of the search task being performed. This does seem to be changing.

Again, this all plays back to the concepts, not keywords, approach to how they're understanding pages as well as the queries users are implementing. To extend this, we need look no further than the knowledge graph

 

The Knowledge Graph Connection

While I am not entirely sold that Hummingbird is really about the ol' KG, (which was also apparently updated recently as well) it is probably is worth looking at. One thing we do know is that this is another more recent commitment from Google in the evolution of the results. And indeed we can see the knowledge base 'cards' showing up in conversational search, a stated goal for the Hummingbird implementation.

Google has become a destination, not just a search engine. One such example we've seen recently include queries such as;

[madonna albums]

b2ap3_thumbnail_Madonna-Albums.jpg

[madonna songs]

b2ap3_thumbnail_Madonna-Songs.jpg

As you click around these results they produce further refinements and plenty of knowledge graph elements. Next let's look at a bunch of examples Stephen Watts shared, using Google conversational search;

b2ap3_thumbnail_KG1.jpg

KG2.jpg

KG3.jpg

Again, while I don't believe there's a direct connection with Hummingbird, it does further highlight the concepts and named entity aspects of the evolution. Not to mention the madness happening in the display changes.

 

What does it mean for SEO?

That's the the big question right? That one's hard to answer. We really don't know enough at this point to really make any assumptions. For those that have already been evolving, getting past keywords, then I doubt a whole lot would need to be changed. On the other hand, if the loss of keyword data recently was a crushing blow, you might have already been behind the times.

Start thinking of things in the form of concepts and semantic baskets, not keywords. In the past you may have created multiple pages to target multiple terms... you know, the ol eHow approach;

  • [how to fix a vacuum cleaner]
  • [how to repair a vacuum cleaner]
  • [how do I fix my vacuum cleaner]


In the past, we might actually target a page to each of those terms. Due to the fact Google was so keyword centric, this type of approach was often quite successful. Given that we could start seeing more implicit predictive results towards potentially synonymous queries, that might not be the case in the near future.

Vacuum-SERP.jpg

When we're creating content for pages we need to get beyond simplified keyword centric approaches. Google sure is. My instinct is that there will be less meat on the bone for overly targeted pages focused on a given term. Changes in query analysis means that your content should be strong on the concepts and related concepts (as well as citations etc) contained on them.

I would also venture to say that those involved in markets that are heavy in mobile, that you will have to also consider how the queries will be formatted, compared to reg desktop style searching. If you read this post (for mobile) that looks at geo-location and behavioural elements, you can start to see where something like Google's Hummingbird, could play a prominent role.

Beyond that, I am not entirely sure how it will affect things beyond targeting and page mapping. It's far too early for that. Many of the same signals used in rankings are still in play. In fact, probably most of them. It is more about the query analysis and by extension, the results returned as far as I can tell so far. I would imagine that semantic mark-up will be more prominent, but I am not convinced that has anything to do with Hummingbird itself.

Early days. As I get more time to think and play with things, I shall post some more. Feel free to drop your own theories in the comments, all are welcomed.

 

And as you go, it bears reading this post from Google on it, there's a TON of mobile phone pics... Coincidence? He he...

Hi my name is Dave and I, am an algo-holic



I am an avid search geek that spends most of his time reading about and playing with search engines. My main passion has always been about the technical side of things from a strong perspective rooted in IR and related technologies.



Also hook up via

Comments

  • Guest
    Jeremy Rivera Saturday, 28 September 2013

    I was noticing that semantically related words were now being highlighted as part of the query on the SERP and thinking how helpful that's going to be for fields with multiple names for the same job. I.e. Attorneys=lawyers, real estate agents=real estate consultants, hairdresser=stylist etc.

  • Guest
    Zach Hoffma Saturday, 28 September 2013

    Great post, I am a big believer Google knows more about where their audience is going and how quickly the audience that used Google may be using other resources than we would know. Mobile is with no question, the biggest evolution in Internet and computer usage since the Internet was created. What we don't know, and Google will not share is how many people are doing what we are calling "natural language search" or flat out just asking friends on Facebook and Twitter.
    There is plenty of market share on the Internet and enough to go around for Google and others. As being in the industry for 10+ years I believe we are seeing signs of potential declining numbers internally at Google that they had never really experienced. It is a challenge to appease Wall Street and business strategy for any organization, apparently Google is not immune to this. And thank you to the teams at Facebook and Yahoo that are speeding this evolution up.

  • Guest
    Elke Greim Saturday, 28 September 2013

    Thank you David for that great value article. Keywords are out - you are totally right and semantic marketing is in. If SEO´s concentrate - like we do - in writting good content and mixing it up with social features they will have a lot work to do in future, too. Many regards from Germany. Elke Greim.

  • Guest
    Chase Billow Saturday, 28 September 2013

    Good article, thanks.

    I'd like to know more about what exactly Hummingbird is (and have the technical training to understand that explanation) but if Singhal is comparing it to Caffeine with 90% of searches affected, I'm guessing it's something fundamental and more infrastructure-related rather than just how search results are displayed - if that makes sense.
    Whatever it is, I'm sure it was necessary to enable a certain change up in the way results are handled and displayed.

    The bottom line is they're getting closer to displaying that one answer to the query, no "searching" needed. I'd like to see some more examples with e-commerce, "transactional" queries instead of informational ones. Those cute little examples they like to give at press conferences are not very helpful and have little to do with how they're going to continue making more and more money.

    They must have a plan for a business model shift bc I'm wondering how they can increase earnings when these Knowledge Graph layouts seem to show less ads. Maybe the price of PPC going up (because more people will be using it) in local, Shopping, etc will offset the loss on informational searches.

  • Guest
    Ravi Monday, 30 September 2013

    Here's a thought, how do you think this will affect Wikipedia? Slowly Google replace it and similar resources as the destination long term. Structured data on websites will get broader covering more elements so Goggle can extract meaning better. Does this seem to be a possibility?

  • Guest
    Debi Davis Monday, 30 September 2013

    In the discussions I've read about Hummingbird there is a lot of emphasis on semantic search, the decreasing importance of keywords, and movement toward accommodating voice search. But, I'm not reading anything about Google taking into account data about the source of the search -- that is, the person; the individual -- and whether or not that influences the search results.

    For example, if I search "fried green tomatoes," will Google recognize that I have visited 40 gardening and recipe sites in the past month, and therefore give me results related to food rather than movies? (Just in case you don't know, Fried Green Tomatoes is the title of a movie.)

    I'm sure this aspect of search is addressed somewhere, but it's not getting much mention related to Hummingbird. How do you think it fits in?

  • Guest
    Ammon Johns Wednesday, 30 October 2013

    Hi Debi. Actually, its something that we mentioned a little in the long live chat we had on Hummingbird and the land of Semantic Search. I even touched on it very briefly in regards to my article on how Hummingbird is kind of the exact opposite of 'long tail' search.

    Hummingbird enables Google to pre-process the query better, applying at least 2 stages of pre-processing, so yes, this can easily include extra steps for contextual lookups too. Contextual obviously includes such things as your location, your previous search history (an essential ingredient of conversational search) and could in future include such things as recent status updates on social media, etc.

    Conversational search in particular pays a huge amount of attention to consecutive searches - such as those "Who is the President of the US?" and "Who is is wife?" examples, where all of the second search is entirely dependant on the context of the prior one.

  • Guest
    Gareth Saturday, 05 October 2013

    Already seeing a few optimizing for hummingbird posts with people talking utter shit, amazing how so few SEOs even try and understand algos properly

  • Guest
    Bhavya Sunday, 06 October 2013

    Hi Dave. Kudos to a post which actually made some sense in the midst of all the crap shooting happening right now when it comes to Hummingbird.

    I have a question to which I would love your thoughts on.

    How does this impact broad based searches. A client's site was ranking well for core terms like "billing machine" and "barcode printer" till the update. After the update, the pages have vanished from rankings.

    Everywhere, I have been reading that HB is more about capturing user intent through long tail keywords and thereby showing results which are more localised, in context and relevant to what she needs at that point of time. How do we extend this update to broad searches like "barcode printer" ?

    One logic I can think about is that the algo keeps a track of subsequent searches people might do for Barcode Printers, for example, "Which Are The Best Barcode Printers?" and thus this impacts the root keyword too. What is your take on this?

Leave your comment

Guest Wednesday, 23 April 2014

Recent Comments

Congrats guys. Keep on doing the same great work.

Alexandra Petean-Nicola
Hi Debi. Actually, its something that we mentioned a little in the long live chat we had on Humming...
Ammon Johns
Hi Dave. Kudos to a post which actually made some sense in the midst of all the crap shooting happen...
Bhavya
Already seeing a few optimizing for hummingbird posts with people talking utter shit, amazing how so...
Gareth
In the discussions I've read about Hummingbird there is a lot of emphasis on semantic search, the de...
Debi Davis
Please update your Flash Player to view content.
Restore Default Settings