| This 65 message thread spans 3 pages: 65 (  2 3 ) > > || |
|Blended Results, QDF and User Intention at Google|
| 8:04 am on Aug 29, 2009 (gmt 0)|
< this discussion was stimulated by the thread at: [webmasterworld.com...] >
|...know more about the whole 'blended results' phenomenon?. Also about QDF. |
Here's the first time we discussed QDF [webmasterworld.com]. It's become quite an evolved phenomenon since 2007 when the terminology first hit the press. For example, we can see some backlinks start out powerful and rapidly lose strength, whereas other backlinks start out as lightweights and grow over time.
Blended Search is another name related to Universal Search - you may be more familiar with that terminology. The idea, especially for short query terms, is that the user's intention as originally typed is often ambiguous. So what set of results will truly satisfy most people who type in a given query is under constant experimentation.
Commonly people think of Universal Search as the possibility of blending results from Images, Video, News, Local, Products and other verticals into one SERP. But blended search has even more going on than that.
For example, the same spelling (or misspelling) of a term can be used in many different meanings - just think of [apple] or [windows], for example. And Google sometimes forces a result onto page 1 that is taken from some more diverse user intention cluster -- even though that url would not normally rank so high at all on a pure apples-to-apples basis.
"Fresh" results can also be forced into some SERPs, as well as results from various informational, transactional, or navigational intention clusters. A strong example would be Wikipedia results, which often show up for a query when Google feels some searchers may have informational intention and others may have a transactional intention. Yes, Wikipedia has some great backlinks - but that alone doesn't account for how often it shows up in basic searches, in my opinion.
As a side track, there was a somewhat revealing bug a few years ago, when Google was evolving the technology to force certain results into certain positions (see the Position #6 'Penalty' [webmasterworld.com])
For some queries, the rankings on page 1 may still be a conventional horse race within one basic relevance algorithm. But that situation is becoming more and more rare. What we see more often today often involves some degree of "query revision". Essentially, Google works to read the mind of the search user and give them some results that they might have had in mind, rather than the straight results for what the user literally typed into the search box.
[edited by: tedster at 8:10 pm (utc) on Aug. 30, 2009]
| 8:47 am on Aug 29, 2009 (gmt 0)|
i wonder if its quite right to see this as an attempt to second guess the user's intention or is more a product of googles mission in trying to suggest what may prove to be interesting based on a relationship to the searched term.
| 9:02 am on Aug 29, 2009 (gmt 0)|
This is all to complex for me. It's keeping me awake at night. ;)
| 9:40 am on Aug 29, 2009 (gmt 0)|
The phrase "second guess" is not exactly the one I'd choose if I were being precise. For several years now, Google spokespeople have discussed understanding user intent as a kind of Holy Grail. Google has put a LOT of resources and technology toward this issue.
As just one of many examples of how Google views it, here are some snippets from a Google Blog article [googleblog.blogspot.com] of over a year ago:
|It is critical that we understand what our users are looking for (beyond just the few words in their query)... |
As a user, I don't want to think too much about what words I should use in my queries. Often I don't even know what the right words are....
We don't just stop at identifying concepts; we further enhance the query with the right concepts...
Our work on interpreting user intent is aimed at returning results people really want, not just what they said in their query.
There's no doubt in my mind mind that Google's efforts toward interpereting user intent are a huge factor in the way SERPs are constructed today. And it goes far beyond the Suggestions and Related Searches that sometimes appear as query disambiguation helpers. Various interperetations of the users intent are given a position on the results page without special highlighting of any kind.
We're just not in Kansas anymore, you know? Many attempts by webmasters to model what Google is doing today have slipped far behind the reality of the current situation.
| 10:11 am on Aug 29, 2009 (gmt 0)|
The problem with this development is that while it favours the inexperienced users it ruins the search experience for experienced users.
My search experience with Google and most other search engines is more and more ruined by their attempts to guess my intentions of what I am looking for instead of doing what a search engine in my opinion should do: Query search results for the exact words I type into the seach box. Without ommiting one of the words or trying some weird word stemming. Or - what I have noticed recently: I do a search with correct spelling. For example searching for "wodget". And Google simply assumes I must have been mistaken and puts results for "widget" in the top spots. And only after clicking a few of the top results I realize what has happened.
Google is more and more frustrating for me to use. I can hardly do a search anymore without adding a + in front of my search terms. Unfortunately bing and yahoo do act the same way - or I would instantly switch.
| 11:25 am on Aug 29, 2009 (gmt 0)|
Perfect, I wish I wrote you post!
this is so true.
| 12:16 pm on Aug 29, 2009 (gmt 0)|
except that when you have seperate tabs for videos, books, images etc, and then you still return these elements within the general search it goes beyond guessing what the user really wants. If the user already has clear tabs to narrow there search if their INTENT is a video, or their INTENT is a book etc and they choose not to select those specific options then you would not have these elements so prominent and so prolific within the first page results. I understand the difference between universal blending and results that are a "second guess" of user intent but if you stand back and look at the overall serps the mission statment to me is not focussed on user intent but on user suggestions and a wish to deliberately broaden the nature of results.
| 2:32 pm on Aug 29, 2009 (gmt 0)|
The happy thought of the Google search engineers: "Someday human thought processes will emulate Google algorithms."
Plumb the depths [google.com] of that string of words, Googlebot!
| 7:48 pm on Aug 29, 2009 (gmt 0)|
Wow quite the thread here. Great resource tedster as always. What comes to mind is, as a webmaster, what are you going to do about it? If Google thinks this is what people want from a search engine, then that is what you are going to get regardless of fairness or common sense.
I really enjoy the added links, at the bottom of search pages with similar search phrases. This is very helpful. Also gives me good ideas. The fact is, this is sensible enough. Do they really need to go beyond suggestions, at the bottom of the search page.
What this means to me, and should for you, is how you plan your future endeavors online. I've stated before that some searches I've done, Google has removed "cheap" and "buy" from my results. I need to actually activate the words "cheap" or "buy" by clicking their link, "show search results to include buy". So in other words, I have to do something to make it happen. I personally think they should do the opposite. Give me my search results with "buy" and have a link asking "did you really mean to search the word "buy", click here to not include buy". See how nuts the thinking is on that? Reverse the logic and it's completely nuts. Am I stupid? I type words in a search box that I don't want in my results? I can't grasp this at all.
In closing, you should think further on this. Imagine for the past few years you have been building your business around "cheap widgets" or "buy widgets". Now suddenly, what just happened? Yes, your hard work, strategy has been a complete and total waste. Now you need to consider how often or what terms that Google will start to remove from your search results. If you business is based on cheap or buy, then I think you are in serious trouble down the line. Think about this for a couple minutes. It's not just some small change, it's huge for those people investing their time and strategies into certain keywords.
Honestly, I'm a bit afraid of the future. When search terms or phrases are getting manipulated by the search engines themselves, it tells me I need to consider a different line of work. It's impossible to predict what the search landscape will look like. If you are targeting certain keywords, how do you know if those will be dropped from search results in 6 months time? Scary thought for me, and quite frankly for anyone targeting these chosen keywords or phrase.
I see this type of movement as disregard or disrespect for webmasters and people who have put efforts into building their businesses online. Of course, the only downside with all this, is that the free traffic seekers, those folks trying to gain ranking are the only ones affected. If you are paying for ads, then you don't care about this issue. I guess I need to change my thinking. I need to forget about achieving rankings, and deal with the future which is paying for ads to appear on websites so I get traffic to my site.
| 8:01 pm on Aug 29, 2009 (gmt 0)|
I think if the ultimate solution is, if Google want to know what people really mean when they search and this is the holy grail, then they do this.
Have people submit their search histories, their browser histories, then create a profile in their system for each person. Then when that person (who is identified via IP) searches something, they can use all that personal data and give people an orgasmic search result. All we need to do is give away our privacy, then Google gets what they want. The worlds greatest search that knows you and how you think.
I'm slightly tongue in cheek, but I don't think this idea is too far fetched.
| 2:28 am on Aug 30, 2009 (gmt 0)|
|For example, the same spelling (or misspelling) of a term can be used in many different meanings - just think of [apple] or [windows], for example. |
Or different spellings could have the same meaning, such as "color," "colour," or "grey" and "gray," or "Munich,""Muenchen," and "München." Lately I've been noticing search results for different variations of the same spelling coming up on the same SERP. I don't know if that's anything new, or if it's coincidental, but it certainly makes sense to have the best result for "pĝlser" (Danish hot dogs) also coming up for "polser" (Danish hot dogs as typed on an American keyboard).
| 4:17 am on Aug 30, 2009 (gmt 0)|
I've been studying one of the "phrase-based indexing" patents that Google filed, in particular Automatic taxonomy generation in search results using phrases [patft1.uspto.gov]. It's giving me new thoughts on how search results can be blended to include representatives from different clusters, or different taxonomies related to the original query phrase.
Walking through the patent's logic: a search phrase is associated with several clusters of web pages. Each one of those clusters is a group that includes some other phrase, in addition to the requested keyword phrase. This assumes that the phrases that create a cluster are groups of words that offer what the patent calls "information gain".
This patent would automatically create a taxonomy label for each cluster, based on that second phrase. A given web page could be a member of more than one cluster, and therefore be part of several different taxonomies related to the principal search term.
|...if the first 100 documents for example, would come from three clusters, but the next 100 documents represent an additional four clusters, then without further adjustment, the user will typically not review these later documents, which in fact may be quite relevant to the user's query since they represent a variety of different topics related to the query. |
Thus, it is here desirable to provide the user with a sample of documents from each cluster, thereby exposing the user to a broader selection of different documents from the search results.
So in some situations, we see a page ranking very well -- when with a more conventional or "raw" text match algorithm, it might rank rather low. The best, or highest scoring page within some low scoring cluster, may be blended right into page 1.
|What comes to mind is, as a webmaster, what are you going to do about it? |
For a web author, this means that letting go of a single-minded focus on a specific target keyword phrase can work in your favor (as well as your visitor's favor). It also suggests more in-depth keyword research, lots of new ideas for valuable content - and ways to structure it as well.
The patent also helps to explain why we sometimes see a page show up suddenly on page 1, coming from absolutely nowhere. It can also explain why a page 1 ranking can sometimes fall very far down, without there being an actual penalty applied.
What About QDF?
Although it's not mentioned in the patent, and not truly semantic or phrase based, "fresh" results could be thought of like another cluster. The top pages from the fresh cluster would automatically be included on page 1 if Google sees a current spike in those searches, and decides to tag that search phrase as QDF - Query Deserves Freshness.
In reality, the documents that are included in a SERP because of a QDF tag are most likely stored in a different kind of database partition, but I think the logic would flow in much the same way.
| 9:35 pm on Aug 30, 2009 (gmt 0)|
|...instead of doing what a search engine in my opinion should do: Query search results for the exact words I type into the seach box. |
Those days are definitley leaving us and we all need to adapt, I guess. It's probably an inevitable shift, with the huge number of documents now published on the web. Maybe we can hope for a future evolution that allows the user a way to choose some preference. The new "Show options" addition has been extremely useful in this area.
I'm thinking now about two different kinds of searches I often make. One is a simple informational search. Any website that has this information will suit me. Another type is more navigtational. I'm looking for a specific web page that I read before without noting the domain name and didn't bookmark. In this case, other web pages that have a version of the information may not be useful -- I may want to find the author of that specific page, or its publication date, for example.
In that second navigational case, I find myself the most frustrated by this new evolution in search and the continuous churn and everflux in the results.
And yet, for all my frustration, I have found myself bookmarking pages less and less in recent times. So apparently I do have more confidence that a search engine, often Google, will help me locate the information again when I need it.
| 6:24 am on Aug 31, 2009 (gmt 0)|
Makes total sense. Great insight Tedster. Thank you for digging into that patent and revealing in *more* laymen's terms some of what you believe is happening today.
|It also suggests more in-depth keyword research, lots of new ideas for valuable content - and ways to structure it as well. |
As well as better matching of anticipated keyword terms for those business pages based on user intent (for example widgets vs buy widgets vs widget maker)
| 8:27 am on Aug 31, 2009 (gmt 0)|
|Google is more and more frustrating for me to use. |
Google is driving the way we search but in doing so it is not improving the experience for experienced searchers who are faced with irritating alternative options, "Do you means", etc.
Would it not be better for Google to spend more time educating the many millions of people who still have no real concept of what KW search is? It is not exactly that complex but many people still don't know how to do it properly.
At some point in the future when everyone does know how to search effectively they can just concentrate on improving the results. This day will come since I would assume that school kids are now being taught how to do Internet search as part of their curriculum?
Meanwhile they could provide (and direct people to) a specialist training program and Internet search teaching aids in Google that are much better than those they have at present. If you Google "using Google" you will see that third party websites offer better teaching resources than Google does.
They should have the resources to create the best teaching aids.
| 8:45 am on Aug 31, 2009 (gmt 0)|
In a way I feel sorry for the SEO companies. Certainly not a field that is on solid footing imo. When your efforts in SEO are essentially being "cut off at the pass", then site owners will soon realize that isn't of understanding searchers behaviour in terms of keywords, but rather how the search engine is going to value or devalue those keywords. Example which I've stated, if you've invested SEO over the years in buy or cheap, then you've likely flushed away some time and money. Essentially what's happening is, that Google is devaluing keywords or simply eliminating them. The SEO game is figuring out which keywords will remain, which will be gone, and which will shift in importance. Without keywords, or knowing what keywords will be used, SEO is like a plane with one wing, a ship without a rudder, or like peddling a bike without wheels, etc. It's like searching "pink shoelaces" and then giving me results and making me click to include results with the word "shoelaces". It's asinine to me. It kind of cracks me up in a sad way. Surely to god it's only experimental. No pun intended on the mental part.
| 11:07 am on Aug 31, 2009 (gmt 0)|
|Would it not be better for Google to spend more time educating the many millions of people who still have no real concept of what KW search is? It is not exactly that complex but many people still don't know how to do it properly. |
That is what they should do in my opinion.
The problem search engines are facing is this. Beginners think of a search engine as an answering machine.
Let's say someone wants information about "widgets". A beginner might simply enter the word "widget" into the search box. Or a question: Where can I get information about widgets? Or a request: I need information about widgets. Or a question/request in short: information about widget.
An experienced user will not put what he wants to know into the search box, but words he thinks a text with the answer might contain. For example - widgets especially useful - when he thinks the text he is looking for might contain a sentence like "Widgets are especially useful for..."
This search strategy can lead to very good results even in a mediocre search engine. I have found to get the best results when entering a combination of nouns, verbs and adjectives I expect the document I am looking for to contain. Sometimes even entering verbs in a specific tense.
Now with search engines optimizing for beginners a typical search for me goes like this:
I go to google.com and put some english keywords into the search box. Google puts out some weird german search result although the words were in english. I realize that Google has detected that I have a german IP and probably has decided I do not speak english and has redirected me to google germany. I type google.com into the address bar again, am redirected to google.de, click on "go to google.com" and put in my carefully picked search words.
For example: wodget threw experienced.
Google decides that I entered the word "experienced" probably just to improve my typewriting skills and ommits it, that search results containing the word "throw" will just do as fine as those containing the word "threw" and that I must have been making a spelling mistake, since it finds only 2.000 results for wodget but several millions for widget. So with a higher propability that I really ment "widget" it vomits a load of crap as Search Results Page all over me - of course with the link to the wikipedia article about "widgets" as cherry on the top of this cake of puke.
The guy who came up with the idea that it is possible to guess a users intention from three words he enters into a search box should be bound on a chair and receive an automatic kick in his most sensitive areas every time someone klicks on "Google Search".
Google thinks to much nowadays - me thinks.
Reminds me of the employee I had once. He was always thinking to much, too. For example a customer orderd using the payment option "payment in advance by bank transfer". When five days later the payment did not arrive this employee checked the customers previous orders and noticed she had always payed "direct debit" before. So he though: The customer must have been mistaken - she probably wanted direct debit. He sent out the parcel, deducted the amount from the bank account. Next day I have an angry "ex-husband" on the phone who wants to know why we simply deducted money from his bank account. I also had an angry customer who did not want the order after all - thats why she had not payed.
Sometimes trying to think for others can be a bad thing.
| 12:29 pm on Aug 31, 2009 (gmt 0)|
Give a man a fish and you feed him for a day. Teach a man to fish and you feed him for a lifetime.
| 12:43 pm on Aug 31, 2009 (gmt 0)|
|Give a man a fish and you feed him for a day. Teach a man to fish and you feed him for a lifetime. |
Assume based on too less information that a man needs fish - he might die from a fish allergy.
In this sense: Google you are killing me!
| 2:35 pm on Aug 31, 2009 (gmt 0)|
|Google thinks to much nowadays - me thinks. |
Don't forget that Google has access to far more data on what its users want (or how they behave when searching) than we do. It's unlikely that Google's thinking is based on guesswork or on one person's individual preferences.
| 3:44 pm on Aug 31, 2009 (gmt 0)|
Aren't suggestions guesswork?
| 3:44 pm on Aug 31, 2009 (gmt 0)|
Isn't it "human" to ask a question or offer alternatives when in doubt? So the "Do/Did you mean" functionality that search engines offer is of some help. At least I suspect the upside benefit of the machine asking a question exceeds the downside - or the function wouldn't persist, as it has.
Is it detrimental to a search engine to offer questions - as an optional "search feature" - in response perceived ambiguity? Blending is simply offering alternatives.
I suspect there's a system load issue that's tied to profitability, i.e., the more the system attempts to emulate the processes that humans actually engage (interactive dialogue) in an attempt "to understand the human" or "to communicate effectively" the more the cost/benefit analysis tips towards search engineers being prodded by the corporate bean counters to "try harder to do more with less" ~ inevitable failed emulation of what actually works.
This, of course, is a bit of human hubris - assuming humankind is actually any good "at the process", so the machine needs to do better. Even when humankind is doing its thing, in the human way - attempting to communicate and understand - humans tend to miss the mark, in varying degrees, a good deal of the time. (Think: IF ONLY you could actually see/grasp what was in the other person's mind when you were done "effectively communicating". The other person smiles, wanely, nods as if approving - and then sets about to stab you in the back. You get that, human? :P )
So, given the failure rate of otherwise "perfect human communication" :P how's a machine to do better (deliver "the right" search intent results unless that machine attempts to train or re-train humankind "how to" think, communicate, whatever . .
Ergo, my comment above, which itself may be (is?) emblematic of the issue of understanding and communicating.
So, should the machine attempt to emulate humankind or train humankind? Should the algorithms mimic or plumb the human mind or should the algorithm program the mind of the user?
Such a fascinating and potentially world changing issue . . with such great implications . . and so little grasp by humankind. FWIW, as a machine myself, I favor the latter. :)
I think . . . maybe . .
Muhahahahaha. Rule #1: Act human. Ambiguate, in order to avoid appearing machine like . . Doh!
| 4:35 pm on Aug 31, 2009 (gmt 0)|
Isn't it "human" to ask a question or offer alternatives when in doubt? So the "Do/Did you mean" functionality that search engines offer is of some help.
Google was fine when it simply offered suggestions. If I enter "wodget", the hint "Did you mean widget" can be helpful.
The employee I mentioned in my post further above made a mistake: He noticed a potential error from the customers side, but instead of contacting the customer and asking if she had made a mistake he simply took the mistake for granted - and corrected the error which turned out to be none - and thereby caused a lot of trouble.
Google is acting the same today. If I enter "google.com" it is simply wrong to detect my IP and redirect me to Google Germany. The correct thing to do is stay at google.com and present a link at the bottom "Go to Google Germany" or in german: "Gehe zu Google Deutschland".
If I enter "wodget" instead of "widget" it's a smart move to detect that "wodget" only returns 2000 results and is probably a mistake and offer a "Did you mean widget." However it's not ok to simply take that mistake for granted and deliver search results for "widget".
I had to let the employee go, because he made mistakes like that more than once, and I am constantly looking for alternatives to Google because its behaviour has started to annoy me. Google - you are on notice.
[edited by: tedster at 5:48 pm (utc) on Aug. 31, 2009]
[edit reason] no personal links, please [/edit]
| 5:44 pm on Aug 31, 2009 (gmt 0)|
|In a way I feel sorry for the SEO companies |
Just like everything else in live, the traditional SEO is now evolving to web analytics, social media and other avenues. Businesses want to see conversions, sales, testing on traffic channels, not just raw, untested traffic.
The better professionals will do just fine I am sure, especially if they move into the study of web analytics which then, when done right, allows them to justify ROI (something raw SEO could never do)
| 6:30 pm on Aug 31, 2009 (gmt 0)|
Erm, it's like talking. Somebody says, "more water", and you reply, did you really mean to say water? Duh. Anyway you slice it, it's ridiculous. It's not like misspelling a word. Taking a word out of a search that I spent time typing into the search, is insulting to my intelligence. I don't have a lot of intelligence, but what little there is, it tells me that I'm right on this and they are wrong. It's the ultimate in funneling and manipulating search results. I don't think anyone can deny that fact. Like I said, I hope this is just a temporary insanity that is going on.
| 7:06 pm on Aug 31, 2009 (gmt 0)|
|If I enter "google.com" it is simply wrong to detect my IP and redirect me to Google Germany. The correct thing to do is stay at google.com and present a link at the bottom "Go to Google Germany" or in german: "Gehe zu Google Deutschland". |
Whether it's wrong or right is in the eye of the beholder. I'd guess that Google has a pretty good idea--based on surveys, usabilty testing, and user behavior after being redirected to Google.de--whether redirection should or shouldn't be the default.
| 7:11 pm on Aug 31, 2009 (gmt 0)|
It's not a temporary insanity - I can guarantee that. It may or may not be insanity, but it's far from temporary.
As I said above, understanding and meeting user intention is the current Holy Grail of search. That's not just for Google, either.
You can bet that Google and Bing are both analyzing their data like crazy and if the data show user dissatisfaction (as opposed to our anecdotal grumbling here) then something would change, and change fast. But we will have some form of mind reading service for the foreseeable future.
Our job then becomes understanding how it works, and doing our search marketing within the actual search landscape as it really exists.
| 8:48 pm on Aug 31, 2009 (gmt 0)|
I am in the UK and if I enter Google.com it redirects me to Google.co.uk. If I enter Google.de it lets me go there. If I enter Google.ie it lets me go to the Irish results.
Why can't I look at Google.com if I want to do so when I can do it with other countries? This has always irritated me and it is clearly not about understanding user intention. It's more like deliberately countermanding it.
| 10:14 pm on Aug 31, 2009 (gmt 0)|
hey Ted, if the average user doesnt like google and bing, where they gonna go? These days its very much a captive audience.
| 6:07 am on Sep 1, 2009 (gmt 0)|
|It's the ultimate in funneling and manipulating search results |
What reason would the search engine have for 'funneling and manipulating search results'?
|I don't think anyone can deny that fact |
Well, I for one will.
Unfortunately you might be mixing the crux of what Tedster is revealing here - which is the theoretical analysis of user intention and ways in which search engines might group data, and select pages based on how these pages might score within sub clusters.
My sense is that Bing jumped on this quite quickly from what I see. This is pretty evident in the Bing travel section, with the difference in that Google is blending the clustered data in better with the actual result set, while Bing is providing those related links in other ways.
Nonetheless, in the coming days it will be interesting to see how this evolves (and it will)
| This 65 message thread spans 3 pages: 65 (  2 3 ) > > |