Deviations from Boolean

booleanstringsBoolean Leave a Comment

Is Boolean (newly) broken? First, let us make it clear what we mean by the question. There are:

  1. Google Boolean search
  2. Google search
  3. LinkedIn Boolean search
  4. LinkedIn search.

Out of the four, two are intentionally “broken,” and one is broken due to bugs. One works fine, but produces unexpected numbers – and lists – of results – something that has started to happen only recently. (Let me know if you need info on which is which.)

Formal Boolean search logic has been intentionally “altered” on both Google and LinkedIn for a long time (the image here is a joke.) Google gives us some control over making a search “strictly” Boolean via quotes and Verbatim. LinkedIn does not.

But Boolean deviation is not a problem if implemented well. On Google, you can benefit from its semantic algorithms when you search as everybody, i.e., to find one good answer. We all “Google” when sourcing and exploring relevant companies, job titles, associations, etc. in the Exploratory part of our jobs. These are short “questions,” easily fitting under Google’s limit of 32 keywords.

What is new on Google, is that if, unlike most humans, you are after getting as many results as possible, you need to play some games and combine results. Rephrasing a query in a formally equal, but differently expressed Boolean Strings generates widely varying results and their numbers. Even a Google Search Engineer found some search strings to return “odd results” in my post.

The reason Google returns such different results to similar queries is that our queries are of a rare, “restrictive,” kind. Google’s attention – both in the ways Engineers develop algorithms and Google AI teaches itself by watching everyone’s reaction to results – is targeted at the vast majority of population that just “Googles” and hopes to see one correct answer. We had an informal chat about this on Recruiting Brainfood last week.

I am guessing with the reasons being:

  1. The growing amount of data in Google’s Index making its fast response harder
  2. A desire to return some results if at all possible and knowing that most want only one or a few results,

Google regularly restricts numbers of results to small percentages of matching results. It picks results rather randomly when you switch or repeat keywords, mainly because it has not been trained on such “weird” requests. It never returns more than 300-400 results (which means that its displayed numbers of results mean nothing. It never did; back in 2010 we still saw up to 1,000, but no longer). However, as of recently, some queries return more (close to 400) or fewer (less than 100) results after seemingly “cosmetic” tweaks. Google stops when it “feels” it has made “enough” effort on your query.

Consequently, silly things like repeating keywords, putting some of them in quotes, or excluding non-existent keywords make Google “grab” different results that sometimes overlap very little – to our advantage. By scraping and combining records, you can beat up all the limits and get lists of over 1,000. It is similar to a spy or a thief scanning a dark room with a flashlight and getting a fuller picture.

Another effect: because Google wants to make you happy, it will trigger its semantic components, interpreting your search if it finds few exact matches. It differs from before in how much effort Google puts into varying your terms. The more keywords you add, the more extensive range of similar words you are likely to see. Google will usually initially try to make it work with one variation, like owner as a synonym for manager. If you exclude that, it will try to vary something else until it feels that you have been served. It used to be more generous with synonyms before 🙂

Quite a different story, LinkedIn personal people search is broken, and you need to know how.

Join me this Thursday for a brand-new class,

Search Is No Longer Boolean,

where I will review the specifics of the Search Universe changes, demonstrate examples, and discuss search strategies.



Leave a Reply

Your email address will not be published. Required fields are marked *