Bing’s Semantic Search, Phonetics and Undocumented Operator

Posted by | November 14, 2011 | Bing, Google, Semantic Search, x-ray search | 16 Comments

I was recently performing some searches on Bing and came across something curious that I had never noticed before.

I’m not exactly sure if what I found is new or simply something I’ve overlooked in the past. I updated Twitter with “Did you know that Bing supports the + query modifier?” on November 10th, wondering if it was something that other people knew about.

I only received a few responses, including a couple from noted sourcing luminaries, and the consensus was that I didn’t find anything because it wasn’t documented anywhere and they could not get it to work.

However, the +/Plus sign does in fact work when searching Bing – just not like it used to in Google.

It’s always a little exciting to think you are one of the first people to stumble across something most people don’t know about, although I won’t get my hopes up that I’m the only person outside of some folks at Microsoft who’s ever figured out that Bing supports the +/Plus sign in searches.

This discovery also led me to proof of Bing leveraging semantic and phonetic search

Bing Search Supports the +/Plus Sign

So I was tinkering around on Bing testing very basic LinkedIn X-Ray searches (more on that later), and here’s my original Bing search of LinkedIn: site:linkedin.com “location Houston” java

Here are the results I found – notice anything odd?

 

 

 

I immediately noticed that the 3rd, 4th, and 5th results highlighted keyword hits of “Coffee.”

My first response was confusion – I could not recall Bing ever trying to so obviously perform semantic search and attempt to guess what I might be looking for by returning results with related terms I didn’t actually search for.

Then I scanned back up the page and noticed something similar to what I see on Google all the time, typically when Google thinks I might have misspelled something:

 

When I clicked on “Do you want results for site:linkedin.com “location Houston java,” this is what I saw:

 

The first thing I noticed was the +/Plus sign.

I could not recall ever seeing it before when searching Bing.

Then I looked at the results, and it was obvious that the +/Plus sign was serving to remove Bing’s attempt at semantic search and only return results with the exact terms I searched for.

No more results mentioning “coffee” when I was searching for Java.

If you think my observation of the +/Plus sign was a fluke, the very next day I was helping one of my associates with a search and noticed he used “HSCM” in an OR statement for a PeopleSoft FSCM position. I had never encountered HSCM before on a resume making reference to anything PeopleSoft SCM related, so I Binged it.

My search was simply PeopleSoft HSCM.

When I saw the results, I noticed the “Including results for peoplesoft hcm,” even though I searched for HSCM.

 

 

In this case, I don’t think Bing was trying to perform semantic search and return a related search term – I think Bing was actually steering me towards a spelling variant that is more common to Bing’s index, perhaps assuming that I misspelled the term in my original search.

When I clicked on “Do you want results for PeopleSoft HSCM,” there were only 31 results, and the +/Plus sign was there, preceding the search string:

 

 

If you try the same search on Google, Google doesn’t give you the benefit of the doubt and simply assumes you misspelled your search term and gives you results for what Google assumes you were searching for.

 

 

How rude.

I know it’s a stretch, but there are some people who actually do know what they are searching for and would rather not have their searches hijacked.

Bing vs. Google

I was a very early adpoter of Google’s search engine (think 1998), preferring it over what most “power searchers” were using back then (think AltaVista).

For many years I was a Google extremist – I used Google search for literally all of my searching needs and never bothered to search using any other Internet search engine except for experimental poking around.

However, not too long ago, after getting frustrated with the junk Google was returning in my LinkedIn searches as well as Google more frequently questioning my humanity by forcing me to jump through CAPTCHA hoops , my experimental poking around with Bing got more serious.

At this time, I use Bing more than I use Google – I’d estimate a 60/40 split.

Part of this is driven by the fact that I find Bing X-Ray searches of LinkedIn are so much “cleaner” and not subject to as much “noise” as Google search results. I also find searching for LinkedIn profile headline phrases in Bing to do a very good job of returning the profile I’m looking for, even if I don’t use the site: command to specifically search LinkedIn.

And of course I love the fact that Bing supports configurable proximity with the NEAR:X search functionality, allowing me to perform feats of magic and semantic search at the sentence level.

I also like the fact that, as I showed above, Bing will by default include your search terms along with results of terms it thinks you might find relevant.

With similar searches, Google just assumes you don’t really know what you were searching for and gives you results of what it thinks you were searching for.

And if you happen to be searching for flights, Bing’s Price Predictor totally rocks!

Unrelated to sourcing and recruiting, I know – but a gem nonetheless!

Bing Searchers Beware of Semantic and Phonetic Search

Now that I am on the lookout for Bing’s semantic search, I’ve noticed that sometimes Bing will slip in semantic search results without giving you the “Including results for ____ / Do you want results for _____” heads-up that lets you know Bing has included results with terms you didn’t actually search for that Bing thinks is related and relevant.

For example, here are the first page search results for a Java search that returns “Coffee” and more interestingly “Coffey” – which means that Bing is not only going semantic by returning words that may have a similar meaning in certain contexts, but also phonetic, returning words that sound similar to the search term.

 

 

If you explore the cached page for the Coffey result, you will notice that there isn’t any mention of Java anywhere, so the only thing I can conclude is that Bing took my search term of Java and leveraged semantics to also search for coffee as well as phonetic variants, such as Coffey.

I know there have to be a few fellow search geeks that find that prospect to be quite interesting. It looks like the folks behind Bing search have been busy!

In any event, the real lesson here is that Bing didn’t give me a heads-up that it decided to also return results with terms I didn’t actually search for.

So, if you’re using Bing to search for anything and you don’t want it taking any liberties with semantic search because you only want results with the exact search terms you used, be sure to add a +/Plus sign to the beginning of your search, like so:

 

 

Related Posts Plugin for WordPress, Blogger...

About Glen Cathey

Glen Cathey is a sourcing and recruiting thought leader with over 16 years of experience working in large staffing agency and global RPO environments (>1,000 recruiters and nearly 100,000 hires annually). Starting out his career as a top producing recruiter, he quickly advanced into senior management roles and now currently serves as the SVP of Strategic Talent Acquisition and Innovation for Kforce, working out of their renowned National Recruiting Center with over 300 recruiters. Often requested to speak on sourcing and recruiting best practices, trends and strategies, Glen has traveled internationally to present at many talent acquisition conferences (5X LinkedIn Talent Connect - U.S. '10, '11, '12, Toronto '12, London '12, 2X Australasian Talent Conference - Sydney & Melbourne '11, '12, 6X SourceCon, 2X TruLondon, 2X HCI) and is regularly requested to present to companies (e.g., PwC, Deloitte, Intel, Booz Allen Hamilton, Citigroup, etc.). This blog is his personal passion and does not represent the views or opinions of anyone other than himself.

  • Lois

    Another great article Glen, thanks for sharing!

  • http://twitter.com/jt1989 John Turnberg

    Excellent find.  Besides adding the +/plus before your search string it appears you can -/minus sign to remove synonym from your results.  This is useful when multiple terms in your string have ‘suggestions’

  • Pingback: Bing and the Plus Sign : Beyond Search()

  • http://twitter.com/braingain Irina Shamaeva

    Hi Glen,

    Actually it seems we knew the plus works on Bing but we have forgotten… Bing’s documentation is sort-of “distributed”. You can find a note from yours truly here back in 2009 http://booleanstrings.ning.com/forum/topics/google-vs-bing-comparison-on and the Bing list including the plus at http://onlinehelp.microsoft.com/en-us/bing/ff808438.aspx

  • http://www.booleanblackbelt.com Glen Cathey

    Thanks Irina – nice find! Unfortunately, Bing’s search operator documentation is either incorrect, incomplete, or perhaps just out of date with regard to the +/plus sign.

    In the documentation you reference, it says the +/plus sign “Finds webpages that contain all the terms that are preceded by the + symbol.”

    However, as I demonstrated, the plus sign is definitely doing something more than simply returning webpages with the terms preceded by the +/plus sign. In my examples, the search terms were present both with and without the +/plus sign, so the +/plus sign wasn’t required to return my search terms.

    Furthermore, as I illustrated, you don’t need to put a +/plus sign directly in front of each search term for it to affect the search term.

    For example, if you search for something simple like Atlanta Java in Bing, you will see “Including results for atlanta coffee. Do you want results for atlanta java?”

    If you select results for “atlanta java”, you will be taken here: http://binged.it/rA0Fra

    You’ll notice the search changes to +Atlanta Java, where the plus sign is preceding Atlanta, not Java, yet the plus sign serves to affect the search for Java, turning off Bing’s attempt at semantic search by returning results of coffee. The documentation you reference doesn’t mention anything about using the +/plus sign to prevent related terms from being returned and/or turning off semantics or phonics.

    As such, it does not appear to be necessary to precede individual search terms with the +/plus sign to return search terms, let alone unaltered search terms.

    And speaking of unaltered search terms, the +/plus sign does not appear to serve to return unaltered or exact terms, or at least it doesn’t in some cases.

    For example, if you search for +car in Bing, you will still get pluralization results for “cars:” http://binged.it/sOhoXU

    With regard to exact terms, as the documentation mentions, the quotation marks only serve to return exact phrases. It does not appear that you can use quotation marks in Bing to return a single exact term as you can in Google.

    For example, a Bing search of “java” atlanta still returns hits with coffee: http://binged.it/uA5dxn

    With regard to Bing’s search documentation, I am not exactly sure if the +/plus sign works consistently to search for and return terms that are usually ignored. I’ve found this particularly challenging to test and get conclusive proof without knowing the list of terms Bing ignores – have you found this list?

    Just as a simple example, if you search Bing for +java auto alpharetta server +and (uncapitalized ‘and’ is mentioned as an ignored stop word) and you click on some cached results, you’ll not find any highlighted hits of “and.” And when I CTRL-F and look for “and,” I typically find it, but can’t tell if that’s because it’s such a common word or if the +/plus sign was actually serving to return results with specific mention of “and.”

    Thoughts?

    Overall, what I was most excited about when writing this post was discovering evidence of Bing’s attempts at and use of semantic and phonic search, as well as the fact that you can use the +/plus sign to turn off semantics and phonics.

    Thanks for digging up that Bing search term guide Irina!

  • Pingback: » Search Engine News Wrap-up Nov 20()

  • http://twitter.com/braingain Irina Shamaeva

    Hi Glen,

    Nice, thanks for investigating! :) The Bing documentations says “all the terms that are preceded by the + symbol.” will be included exactly, i.e. you do not need to put the plus sign in front of each word. 

    In your example with +car I believe each page still has the word “car”, though it’s true that “cars” are also in bold. If you find an example where the + doesn’t force using all of the exact words I’d be convinced! (“and” is too special of a word, I think). 

    What may be new in Bing is, when it decides to look for synonyms and phonetic similarities, it offers the user to include the plus in front of all of the words, that, per its documentation dating a few years back, will not modify them. 

    It’s hard to say how much effort Bing has been putting in synonyms and phonetic similarities – obviously, enough at this point to offer ignoring its suggestions using the +. While Google explicitly says that it searches for synonyms I haven’t seen that mentioned in Bing’s help. I wish they’d update Bing documentation and keep it in one place!

    I also think that quotations marks around a word does bring up pages with the exact word, as it does on Google. It highlights “coffee” in your java example, but the pages found do contain the word “java”, as far as I can tell. It seems it’s a matter of highlighting as well.

    In conclusion, for now, I am still convinced that the + is documented and works as documented, and so do the quotation marks. http://onlinehelp.microsoft.com/en-us/bing/ff808438.aspx

    Looking forward to further discussion! :)

    -Irina

  • http://www.booleanblackbelt.com Glen Cathey

    Irina,
    Challenge accepted. ;-)

    The Bing documentation statement that the +/plus sign “finds webpages that contain all the terms that are preceded by the + symbol” is open to interpretation.

    I don’t interpret that statement to mean that you only need one + sign in front of the entire query to affect all of the search terms. I’d love for someone from the Bing team at Microsoft to settle this one. :-)

    To throw a wrench into the theory that all of the terms must be preceded by a single +/plus sign at the beginning of the query, take a look at these searches and their results in this order, and tell me if you notice anything before you read any further.

    atlanta house server java – http://binged.it/t5KnMX

    +atlanta house server java – http://binged.it/sDjHEk

    atlanta house server +java – http://binged.it/sRPu66

    For the first search, Bing appears to want to look for ‘home’ in addition to ‘house.”

    If you click on “Do you want results for atlanta house server java?” you get search #2.

    Just to experiment, I then decided to remove the + from in front of the string and put it only on the last search term to see what would happen (search #3).

    What I notice is that the 2nd and 3rd searches seem to produce the same results (at least the first 3 pages I checked are identical, as is the estimated total # of results, fwiw), and that the addition of the + sign, regardless of whether it precedes ‘atlanta’ and the beginning of the string or ‘java’ at the end of the string, serves to turn off semantics/synonyms for ‘house,’ the 2nd term in the query. Interesting.

    So the +/plus sign does not have to precede the entire string, as your interpretation of Bing’s search documentation suggests, and it can actually work backwards, affecting preceding terms.

    With regard to quotes, I tried searching for a deliberate misspelling using quotation marks – “coffeee.”

    When you search Bing for “coffeee,” here are the results you get: http://binged.it/rzvkkM

    The issue with those results is that the vast majority are for “coffee.” If you click on some random results, use CTRL-F or check the cached page, you’ll find that most don’t have any mention of “coffeee.”

    That means that the the presence of the quotation marks didn’t serve to do much of anything to return the exact search term.

    If you decided to click on the the familiar “Including results for coffee. Do you want results for “coffeee”?” and you click that you only want results for “coffeee,” here is what you get: http://binged.it/rQe0cd

    First, you should notice that the query changes to +”coffeee.”

    However, even with the +/plus sign AND the quotation marks, you will still get some results that do not mention “coffeee.” With some random checking, I found 6 results (incl. Wikipedia on page 1) that have no mention of ‘coffeee’ Here’s one from page 2: http://bit.ly/vRAoga

    If you try the same thing on Google, you only get results with “coffeee” – http://bit.ly/tvlID8

    This is fun. :-)

    Thoughts?

  • http://twitter.com/braingain Irina Shamaeva

    OK, Glen, sorry, but your first 3 strings of examples can’t serve as a proof that: 

    *** Bing will include all terms following the plus exactly which Bing’s documentation states.The similarity of the last two strings of results *don’t* prove it. Would you agree? (Maybe sometimes it works backwards; that’s good too.)atlanta house server java – http://binged.it/t5KnMX+atlanta house server java – http://binged.it/sDjHEkatlanta house server +java – http://binged.it/sRPu66

    I think I do see Bing breaking its own rules in the example with “coffeee”. It should be the same as +coffeee.

    However, it is fine for a search engine to show some pages that do not include the given word. As an example, a page is relevant if there’s a link to it from another with the given text included. With your example, take a look at this page :) http://statsie.com/marleycoffee.com 

    At the conclusion, I suggest that you search on Google for 
    “coffeee” marleycoffee:
    http://bit.ly/vff7vU

    Your turn! :)

  • http://www.booleanblackbelt.com Glen Cathey

    Irina,
    Ah, but I actually do believe that the identical results for [+atlanta house server java] and [atlanta house server +java] proves that the +/plus sign doesn’t need to be at the beginning of the string. Without the +/plus sign in front of Java or Atlanta, Bing will return results of “home” instead of “house.” Adding the +/plus sign at the beginning of the string or the end produces the same affect on the search and the same results.

    I’m not sure how far you checked, but I checked out to page 25 with both of those searches – the results are identical, not just similar. I’ve done it with other searches as well – these aren’t fluke results.

    I’ve sunk some time into my investigative work and testing – now the burden of disproving it lies with you, with specific examples. Although I am not sure how much more work on either of or parts should be sunk into getting to the bottom of Bing’s (obviously inaccurate and/or outdated) documentation :-)

    While I agree it may be fine for a search engine to show some pages that don’t include the search terms I specify, it is NOT fine for a search engine to show some pages that don’t include the search terms I specify if I use query modifiers that are suposed to return only the terms I specify (e.g., +”coffee”)

    With regard to Googling for Coffeee MarleyCoffee, I got these results, all with the misspellled “coffeee” http://bit.ly/u63NvV – not sure what you were referencing?

    And now for something interesting:

    +atlanta house server java+

    Results http://binged.it/uQeBeF

    atlanta house server +java+

    Results http://binged.it/uqPZRo

    Identical results for both, and they do find pages with “Java+” – I checked a few cached pages and used CTRL-F to find “Java+”, so at least the +/plus sign is allowing you to search specifically for the +.

  • http://twitter.com/braingain Irina Shamaeva

    I’d hate to go in rounds… I wonder if we have different assumptions.

    Let me try again. 

    1. Bing documentation:

    “If you put the plus sign within your string, all words after that plus will be searched for exactly. ”

    A=” you put the plus sign within your string”
    B=”all words after that plus will be searched for exactly”

    The rule in Bing’s documentation about the plus says: A->B

    To prove that it is WRONG, you would need to find a search where the plus is used but some of the words that follow are not found exactly., i.e.

    A-> (NOT B)

    What you are offering is this:

    A=”you put the plus sign within your string”
    B=”all words after that plus will be searched for exactly”
    C =”all words before that plus will be searched for exactly as well”

    Glen says: A->B+C

    –> OK, maybe (I don’t know), great, so what? It is not against the Bing “plus” documentation.

    2, The very first result in Google’s search that you also ran, http://bit.ly/u63NvV, DOESN’T have the word “coffeee” with 3 “e”s that we had asked for explicitly by providing the quotation marks (for me the first result is www.marleycoffee.com)

    -Irina
    The search for +atlanta house server java+ and the other one show interesting results indeed! Oh my!

  • http://www.booleanblackbelt.com Glen Cathey

    It appears we’ll have to respectfully agree to disagree. :-)

    There is no assumption – the exact Bing search documentation phrase is that the Plus sign “finds webpages that contain all the terms that are preceded by the + symbol,” which means that the terms must be “preceded by the + symbol.”

    To me, the critical word is “preceded.”

    If a term is NOT preceded by a + symbol, it doesn’t necessarily need to be returned as specified – Bing is free to return synonyms it believes might be related and relevant, as it does with my example of [atlanta house server java] – where Bing returns hits of “home” for “house.”

    However, if a term is “preceded by the + symbol,” it should come back as specified, without Bing offering suggestions.

    Trouble is, in a string such as this: [atlanta house server +java], the ONLY term preceded by a + symbol is java.

    Because “house” isn’t preceded by a + symbol (nothing but the last term of ‘Java’ is), Bing should be able to return synonyms such as “home,” but it doesn’t, apparently because of the + on a term that succeeds “house” in the string on a later term.

    For me, it comes down to the definition of “preceed,” which is to “come before in order or position.”

    In the [atlanta house server +java] example, the + symbol succeeds the affected search term (house), which is the reverse of the Bing documentation statement of “terms that are preceded by the + symbol.”

    Ultimately, what my testing has found is strange in that it makes no logical sense to be able to return 4 specific search terms by placing a + sign on only the last of the 4 – and not even the one you are trying to specifically return with no variants. I am not sure who would ever think that the + symbol would or should work on terms it succeeds in the string. Leave it to me to waste my time discovering this kind of stuff.

    The fact that all 4 of these searches return the exact same results…

    +atlanta house server java
    atlanta +house server java
    atlanta house +server java
    atlanta house server +java

    …where the only thing the + symbol seems to be affecting when compared to [atlanta house server java] is the 2nd term in the string (house), would suggest that it would be more accurate for Bing’s documentation to say that “preceding any term with a + symbol will find webpages that contain all of the exact terms in the query.”

    I think the real issue here is that Bing’s documentation is somewhat ambiguous, and certainly not specific. Who knows – maybe I am the only person to ever want Bing to be more precise in their explanation of their search operators and modifiers.

    Thank you Irina – I always appreciate a different perspective, and certainly respect yours. I actually prefer disagreement over agreement because I’ve found it prompts critical thinking, which we can never have enough of.

  • Pingback: Search Engine News Wrap-up Nov 20 | Domain Buddy()

  • Pingback: What’s Up With Bing Changing My Searches? | SEER Interactive()

  • Pingback: What's Up With Bing Changing My Searches? | SEER Interactive()

  • Pingback: Bing’s Semantic Search, Phonetics and Undocumented Operator | 4sct()