If you look at a typical page that shows up after you perform a search at one of the major commercial search engines, you’ll see that those search result pages don’t differ too much from each other.
Some sets of search results do include news, images, maps, and other results that go beyond just a list of web pages that may contain the keywords used in a search.
But, how interested would you be in entering the address of a web page and seeing related search queries for that page, or related people or places or other pages?
Inversion Searches Showing Related Search Queries
This kind of search, referred to as an “inversion search,” by some Microsoft inventors, is the topic of a new patent application from the Washington-based search provider.
Information about related queries from this inversion search, combined with data about how people might click through those queries and spend time on pages visited through an inversion search, might be used by the search engine to improve search results and rerank those results.
The patent filing also describes how web page owners could elect to show related queries about their pages, upon their pages, along with advertisements and share revenue from those advertisements with Microsoft.
The patent application is:
Related Search Queries for a Webpage and their Applications
Invented by Krishna Chaitanya Gade, Srinath Reddy, Nicholas Eric Craswell, Amit Prakash, and Hugh Evan Williams
Assigned to Microsoft
US Patent Application 20080235187
Published September 25, 2008
Filed: March 23, 2007
Abstract
An inversion of the basic format of searching is provided herein. Instead of receiving a search query and providing web page results, a search engine receives a web page identifier as search input from an end-user, determines related search queries for the associated web page, and provides the related search queries to the end-user issuing the search.
Related search queries for web pages may also be used to refine search engines performing the basic form of searching by facilitating the determination of web pages to index and the ranking of web pages as search results to user queries. Additionally, related search queries may be used in advertising revenue generation and sharing.
The search engine might collect information about related search queries by extracting keywords from the content of a target web page and generating related search queries based on the extracted keywords.
Query terms might also be taken from historical search information taken from search engine query logs, including the targeted web page, or from the search engine’s cache file, or from previous inversion searches for the page.
Those related search queries might be presented based on rankings determined from the relevance of the related search queries to the web page and the popularity of the related search queries based upon the user’s historical search information.
Keywords might be extracted from web pages by looking at anchor text, page titles, and text appearing in the page’s content. Words found might be stemmed, to find the roots of the words and vary based upon those roots.
Stop words, that appear frequently and aren’t important to the page’s content may be removed. Words that don’t appear too frequently upon a page (below a certain threshold) may also be not be returned as a related keyword phrase.
Term frequency and inverse document frequency (TF/IDF) techniques may also be used to try to score words and phrases against each other on a page, to find the terms with the highest scores, and to identify them as keywords.
When a search engine ranks web pages, it may try to match the query terms with words “occur in several parts of web pages, such as the anchor text, title, body, and URL string.” The weights for those different parts may be “tuned manually” or through “machine learning techniques.”
The patent filing tells us that these rankings could be adjusted by looking at both related search queries and extracted keywords, and the user interactions with those related queries.
A high frequency of users selecting a particularly related search query for a web page may be viewed as empirical evidence that the web page should be considered highly relevant for the selected related search query.
Accordingly, web pages may be given higher weighting for search queries matching related search queries having a high frequency of user selection as evidenced by historical inversion search information.
Conclusion
There may be a few different ways that an inversion search might show up in search results, such as a checkbox or some other way that searchers could select to perform a “related” search, or the use of a special search operator that would be used like this – related:www.example.com, or if a searcher entered a URL for a page into the search box, the search engine might automatically display related search queries.
Inversion search could be a pretty informative tool for searchers, giving them the chance to explore concepts found on pages, and search results for those concepts or related keyword phrases.
I can envision site owners who might occupy the same or similar markets spending a lot of time exploring related keyword phrases for pages on competitors’ sites. How much of an impact might inversion search have on the development of new content for their sites?
This sounds like great fun. With this search would work in both directions. Perhaps you would like to verify information on a page, you could simply invert the search and find related pages to confirm or refute the evidence.
I recently read that to optimize your page on a regular basis, you should add new content frequently. How frequently? I have a full time job and waitress at night 3 times a week, and I dont’ always have time to add to my blog. My current schedule allows me to post 2 times a week – do you think that this if often enough?
Actually sounds like a great tool for web marketers, SEO professionals and keyword researchers more than anything, but that is not a bad thing.
@ agentmulder412
If you are strapped for time, you may want to consider bringing some friends over for group writing sessions (You supply the food).
You can write tons of blog posts and load them in by distribution date so that you have new posts coming every few days. You will cut your blogging time in half and beat burn out.
Its great to see some new technology for search coming out. I’m not as up to date as all of you probably are, but how would this integrate with the semantic web? Or is this just a version of it?
“inversion search” .. I never heard of it before but thanks to you and this article.
Hi Robert,
One of the things that I like best about this inversion search is that it does provide an opportunity to quickly find those kinds of sources, so that you can make some informed decisioins on information that you locate on the web. And it does sound like fun. I’d like to see it developed.
Hi agentmulder
There are no fast set quidelines for adding or updating content to a web site. If you blog about a specific topic, and it’s one where there is a lot of discussion on that topic, updating to become part of the conversations about that topic is one strategy that you might want to consider. But, that isn’t really necessary. Blogging a couple of times a week instead of everyday probably shouldn’t have any impact on how well or poorly a search engine ranks the pages of your blog. Now if you blog only once ever few months, that might.
Hi People Finder,
I do think that this would be a helpful tool for site owners and people involved in internet marketing. I think that searchers would find it pretty helpful too, especially if they are exploring a topic that they don’t know much about.
Hi Ben,
Some nice blogging tips. 🙂
It is exciting to see approaches like this developing.
As for the semantic web, if we consider that a way of empowering computing systems to more easily understand aspects of web pages, and information that appears upon those, the patent filing does describe how user interactions with suggested related queries from inversion search may provide the search engine with additional information on how to rank pages for the related query terms involved. But that isn’t getting additional meaning from the contents of the pages themselves, but rather how people use and view those pages.
Hi Charles,
Thanks. It’s my first time seeing that specific term (inversion search) from one of the search engines, too. You’re welcome.
Bill,
That is true. It would be especially helpful for researchers who are looking into topics ( especially professional and technical areas ) where they are unfamiliar with the specialized lingo that goes with the topic at hand.
William
Well, with Google offering outdated info in their SERP’s I think Microsoft will be wanting to avoid a similar scenario. Or at least offer a better way to verify “breaking news”.
I think it would be great fun.
Here i would like to clarify some information about html codes like “AND” – AND
here Google recognizes this code correctly but others like yahoo, live and cuil does not recognizes as AND instead this search for same query “AND”.
This makes some confusion that how accurate these search engines?
so which search engine i have to go for?
Hi, i would like to clarify one thing about using html codes,
“AND” – this actually represents “”AND”
when we use this query in SE’s google.com recognizes it as AND and displays some search results, but in other SE’s like Yhaoo, live and cuil search for exactly for the same phrase and odes not recognize as AND. so here the thing is which SE provide correct result?
I’m not sure that I understand your question, smuralii.
Are you pointing out the use of “and” as a boolean operator in a search query performed as a search that would include multiple term (like “OR” could also be), as opposed to a search term?
Thanks.
WRT How much of an impact might inversion search have on the development of new content for their sites?, offering inverse search capability to forward searchers would greatly improve geographical and statistical data (ecommerce and offline) regarding everything from new home sales and repairs, through top searches overall.
WRT “The patent filing also describes how web page owners could elect to show related search queries about their pages, upon their pages, along with advertisements and share revenue from those advertisements with Microsoft,” I don’t see the “sharing revenue” part happening; I see private inverse search queries shown in forward search results. The revenue is in businesses paying per response [to consumer queries]. Another valuable attribute is that inverse searches are immune to spam and unwanted clicks and click fraud.
WRT “Term frequency and inverse document frequency (TF/IDF) techniques may also be used to try to score words and phrases against each other on a page, to find the terms with the highest scores, to identify them as keywords,” this, IMO, is not necessary because natural-language inverse searches are semantic searches.
Combining inverse search with forward search is the semantic web.
Hi Joe,
There is some potential behind what’s described in the patent to make a web where we understand better the meanings of content found on web pages, and how they may be connected to other pages. I’m not sure that I agree with every approach described in the patent, such as the use or need for TF/IDF, but it’s possible that if the process described in this patent is developed, that may not be part of what we finally see.
Thanks for responding, Bill. I’ve communicated with MS regarding inverse searches some time ago. Another issue I see is being able to “inverse search forward searches;” Taking that a step further, first the results a forward search engine delivers would have to be precisely relevant, using keywords and phrases, which we all know right now doesn’t happen most of the time. Second would be allowing queries to be crawled, which we all know would erode privacy, and likely eliminate it. Third is allowing responses (true results) to be crawled, in which case transactions would be publicized, which doesn’t lend itself to ecommerce because final transactions are normally private, and, personally, I don’t see any value in making them public knowledge. I look forward to your next communication. Contact me anytime you like.
Hi Joe,
You’re welcome. I want to thank you for your thoughts and experiences in this area. There do seem to be a lot of issues involved in implementing inverse searches, including privacy concerns. Not sure if Microsoft is going to move forward in providing these types of search results, and how they might address the concerns that you’ve raised, but it will be interesting to see what they might try to do.
When you get a chance, perhaps you could register at inversearch.com and let me know your opinion. Inverse searches private, so no concern there. You can send feedback via the site, plus I’ll email you.
Hi Joe,
I’ll do that. Thanks for the invite.