Paid Search Results and Query Expansion using Synonyms and Related Concepts

Search for “cheap cars” at Yahoo in the future, and you might see web search results and paid search results for terms like “job searches” or “bicycles” in the future, according to a recently published Yahoo patent application.

If you’ve been keeping a close eye on Google search results lately, you’ve possibly noticed that sometimes when you perform a search at Google that the search engine might broaden the search results that you see to include synonyms for one or more of the terms that you used for your search.

I wrote a post about that, Google Synonyms Update, in which I pointed to a couple of patent filings that Google made which described a couple of different ways that Google might come up with synonyms for search terms. In the comments section of the post, a couple of people asked what kind of implications this query expansion might have for sponsored search results.

Good question.

The Google patents on synonyms were quiet on paid search, and what the patent filings might mean for them, but chances are that if Google expanded a search to include synonyms, that it might choose between sponsored ads for the original query and the additional query terms.

For example, a search for [Ft Worth web design] might be expanded to include search results for [Fort Worth web design], and sponsored ads shown in that search result might be shown for both the “ft. worth” search and the “fort worth” search.

Google told us in an Official Google Blog post about this new inclusion of synonyms that they would highlight the synonyms in search results. Would they also highlight the synonyms in sponsored ads, if those appear there?

I just did a search for [ft. worth web design], and it appears that they do highlight synonyms when they appear in paid search results, like in the following image where the term “Fort Worth” is highlighted in the top sponsored search section with the yellow background:

Google search results showing 'Fort Worth' highlighted on a search for 'Ft. Worth.'

Is the term highlighted because Google is showing a choice of advertisements for either [Fort Worth web design] or [Ft Worth web design], after expanding the organic web search results?

Interestingly, Yahoo just published a patent application that explores some similar territory, on expanding search results to include documents that might include words that are synonyms when used in a particular context. The Yahoo patent also looks at user data to decide when to show pages that might involve much broader related concepts as well.

Also interesting is that the inventors listed on the Yahoo patent filing have worked on the paid, or sponsored side of search for Yahoo in the past, and the patent filing explicitly includes paid search results.

The patent application is:

Predicting Selection Rates of a Document Using Click-Based Translation Dictionaries
Invented by Rukmini Iyer and Hema Raghavan
Assigned to Yahoo
US Patent Application 20100017262
Published January 21, 2010
Filed: July 18, 2008

The following paragraph from the patent filing gives us a hint at how Yahoo may broaden the search results they see beyond the methods that Google has described in their synonym patent applications:

In an embodiment, although claimed subject matter is not limited in this respect, a method includes automatically constructing probabilistic translation dictionaries from click-through information. Such translation dictionaries may include a database and/or data tables, for example. Translation dictionaries may include word synonyms as well as words and/or phrases that include one or more meanings that may be related to other words and/or phrases.

For example, a translation dictionary may include the phrase “cheap cars”, which may be related to other words or phrases that likely have a meaning corresponding to inexpensive automobiles, such as “used cars”, “compact cars”, “Kia”, “Hyundai”, and so on.

Continuing with the example, such a translation dictionary may also relate “cheap cars” to “job searches” or “bicycles”, since a user entering the query “cheap cars” may be unemployed, and interested in finding a job. Or such a user may have little money so that a bicycle may offer a good alternative to a car. Constructing such translation dictionaries will be described in detail below.

While click-through information may be one source of data from searchers that Yahoo might use, it’s possible that they might look at other information about interactions between searchers and the search engine. While the above paragraph refers to a “translation dictionary,” the Yahoo patent also describes something they call a “probabilistic model” that sounds, in some ways, like the statisical language models that Google referred to in their patents.

That model attempts to predict the probability that someone will choose a certain page, or ad, or other kind of result based upon past search result selections or click-through-rates of ads, and show results and advertising that match those predictions.

When Google or Yahoo expand query results to include synonyms for words in an original query, it appears that they may also consider doing so for the advertisements that are shown with those results.

And Yahoo may even help you find a bicycle, when you can’t afford a car.

Share

29 thoughts on “Paid Search Results and Query Expansion using Synonyms and Related Concepts”

  1. Your last comment “help you find a bicycle” suggests one might get more search results in the future, rather than less, I don’t mind a few million more matches as long as the first few pages are improved.

    I’m a bit surprised that Google’s staff of PhDs had not already tied up all of the search concepts.

    Hopefully these innovations will just effect search entry interpretation and not page optimization for spiders. Thought I had a feel for Key Word In Context (KWIC) then came along Latent Semantic Indexing…arrg.

  2. I hope this new patent will really be of big help to businesses. This is like expanding the horizon. I really hope it is for the best.

  3. This looks like it will have quite an impact on business that targeting specific long tail search terms, this could be +ve or -ve depending on how its implemented. It will also go some way to killing off micro sites that are set up to target specific terms (in other words people profiting from CPA networks and the like).

  4. Hi Jake,

    If Yahoo adopted this in the future, it probably would result in more search results. But, I’m not sure how I would take it if I searched for “cheap cars” and Yahoo showed me ads and organic search results for bicycles.

    Not quite sure how this will impact optimization of pages quite yet. This is a statistical probabilistic model based upon user data from sources such as search engine query logs, rather than anything to do with LSI. I don’t believe that any LSI methods were actually helpful in optimizing pages as it is.

    As for Google, I suspect there might be some other query expansion models that they’ve been exploring that might come up with connections between concepts in a manner that might be similar to what Yahoo’s researchers suggest here.

  5. Hi Andrew,

    Expanding some queries based upon synonyms and related concepts has the potential to get searchers to see pages that they may have intended to see, even if they didn’t quite use the “right” query terms in their search. I’m not so sure how advertisers might feel if they are bidding on exact match phrases, and their ads are showing up for other terms that a Google or Yahoo might think are “synonyms” or “related terms.”

    Might that be happening? I’m not sure, but it’s possible.

    It is expanding the horizon. I don’t know if there’s an easy way of telling how helpful it might be for businessses, though.

  6. Hi Jimmy,

    Good points.

    I see the potential for the search engines to try to fill up search results with higher quality pages that include synonyms and related terms, rather than trying to find pages where all of the words on a page match all of the words in a searcher’s query. That could impact the way that some people use micro sites.

  7. To me this sounds like it could be very helpful for sites that use a more natural approach to SEO, using a variety of keywords in their backlinks. Suddenly they will find themselves popping up on even more search terms that they won’t have done previously.

  8. I am wondering if this is in response to Google’s acquisition of the Orion algorithm (2006) that also did conceptual extensions of search phrases to pull in potentially related topics. Both would be supporting a “berrypicking” approach to knowledge acquisition. (http://www.gseis.ucla.edu/faculty/bates/berrypicking.html) Berrypicking observers that searcher starts out with a focused information need and can be drawn into alternative directions along the way. Professor Bates does stipulate that the searcher will always return to their original information quest.

  9. Hi Marianne,

    Thanks for pointing out the possible relation to the Orion algorithm, and the link to the page on Berrypicking.

    The original press release from Ori Allon on the Orion algorithm included two parts – expanded snippets, or text extracts, which shows you the information that you might be looking for in search results, and results for related words or phrases and concepts. For the press release:

    Take a search such as the American Revolution as an example of how the system works. Orion would bring up results with extracts containing this phrase. But it would also give results for American History, George Washington, American Revolutionary War, Declaration of Independence, Boston Tea Party and more. You obtain much more valuable information from every search.

    One thing I thought was very interesting from this patent application is that the inventors have been primarily focusing upon paid search at Yahoo, and there’s a fair amount of language about advertising and sponsored search in the patent filing, unlike Google’s patent applications on synonyms.

  10. Hi Jake,

    I’m not sure that would make any difference. This patent is aimed at helping searchers find sites that might be relevant for the the intent behind their queries.

    To use one of their examples, if I search for “cheap cars,” I don’t necessarily want to see lots and lots of pages that actually use the words “cheap cars.” Instead, I want to see pages about affordable cars, even if the words “cheap cars” don’t show up on those pages, and even if the anchor text “cheap cars,” and other possible variations of the phrase aren’t included in anchor text on other pages pointing to those pages.

  11. I am not liking the idea of what Yahoo will be doing with their search engine. If I am looking for “cheap cars”, I want to see cheap cars. I don’t mind different keywords showing such as “inexpensive automobiles”.

    -Kai Lo
    Twitter @lomak1985

  12. Hi Kai,

    I’m a little concerned about some of the “related” concepts too. Showing results for synonyms sounds fine, but I’m not sure that the examples of search results for jobs or bicycles for someone searching for “cheap cars” really match the intent of a searcher well.

  13. I’m interesting to know how they search engines will give the advertisers the option to opt in of out of this new functionality.

    I have a VERY niche self help product, it doesn’t make much, but I do continue to market it via the paid listings. (This is the product that got me into Internet marketing and it helps people so I continue to market it!). Because it is so niche I’ve now got a really tight set of keywords. And this new advertising could just sink it, and that would be such a shame.

    I will be looking out for more news on the subject!

  14. Hi Michelle,

    I haven’t seen much discussion on the Web, or from the search engines on paid search and query expansion based upon synonyms, but I understand your concern.

    If you’ve set up exact match keywords to trigger your ads, and Google expands a query to include your keyword term or phrase, should they report upon what the original term was, and the fact that they expanded the search query terms? I know that I would want to know.

  15. Hi Bill, good reply.

    I have seen a few tremors of work, restructuring on a couple of blackhat forums because of this change.

  16. I would rather see the ‘related’ concepts separated from the exact concepts instead of being blended together. If this is the case, then good for Yahoo! But I agree with Kai in the fact that if I search specifically using adjectives, I want those results to come up first. Also, how will this effect SEO? Will firms have to take out a thesaurus and go through all of the possible synonyms of their targeted key words just to stay where they are? And if they don’t, will they drop down even if they target specific words to a greater degree? I guess constant evolution is the name of the game here.

  17. I think that its all a bit daunting at the moment with where things are going but it will work out well. Look at your mobile phone, if you were to take someone who had only got a landline 20 years ago and tell them that their phone would send messages, access the web, have tracking features etc I feel that they would be put off by the idea thinking its too much. The same concept could apply to the search results page today, who knows what it will look like in 20 years but so long as it does not change overnight we will all be fine.

  18. Hi Jimmy,

    It’s interesting seeing how and where ideas spread. And as you note, how people might react to changes differently over time. I think there are people still put off by many of the things that can be done with a phone today.

    Search will definitely look very different in 20 years. It may look very different in 2-3 years.

  19. Hi Web Squad,

    I agree – I’d rather see related terms offered as query refinements (like the “did you mean XXXX” or “Try also XXXX”).

    It’s always been smart to identify synonyms for candidate keywords, and attempt to choose wisely, and then find ways to include synonyms on pages too, if possible. I’m not sure that there’s any set of rules to follow that will fit in all circumstances. I suspect that it’s often going to be on an individual keyword/keyphrase basis.

    For instance, most people may use “Fort Worth” referring to the City in Texas, and most people may use “Ft. Pierce” referring to the City in Florida. Using both “Ft.” and “Fort” on pages for each likely won’t make any difference at all if Google decides to expand a query to use the synonym. Optimizing as well as possible for the best choice is a better bet.

  20. It does sound a bit similar to LSI doesn’t it. But agree with Jake, it’s always been a good strategy to include a wide variety of related keywords both in on-page & IBLs.

  21. Hi Bullaman,

    I hate bringing LSI into any conversation about SEO, because it’s so often misunderstood and misapplied. :)

    Using synonyms and related terms in copy and in anchor text is more common sense and good practice rather than anything else, and it potentially broadens the terms that a page might be found for. But this patent filing isn’t suggesting that people writing copy use synonyms or related words in their content; rather than the search engine might explore user data to see if there might be some relationships between the usage of words in certain contexts that may make them more likely to be synonyms.

  22. Won’t expanding a query to use synonyms lead to a greater concentration of sites that show up in the first results in Google? And if it does, do you see this as a positive or negative thing? I’ll wait to hear some other people’s opinions before I chime in with my own.

  23. Hi Koby,

    Expanding queries to include synonyms creates the possibility of a wider range of results that might show up in the top results at Google.

    If the synonyms are good matches, then it could possibly be a good thing for people searching for information, as well as site owners who offer what a searcher is looking for, but didn’t use the same words on their pages that the searcher used to find them in his or her query.

  24. Hi Frederik,

    It’s hard to tell how effective this use of synonym matching might be in paid search, or even in organic search. I would expect though, that there’s probably a great deal of testing going on by the search engines. If they do expand broad match in this way, and it results in few clicks or conversions, that could potentially cause advertisers to consider other ways of advertising.

  25. The synonyms are great. We don’t have to focus on 2 words but 1 will do the job. You don’t have to make a site foto photography and 1 for photographer. In google webmaster help program you can see the synonims.

  26. Hi Arvid,

    I agree, but with some level of caution. When you’re paying on a click-by-click basis, it always helps to pay attention to things like this carefully.

Comments are closed.