Replies to This Discussion

Permalink Reply by Simon Townsend on January 4, 2009 at 9:33pm

Google does have a wild card feature which is actually more useful than most wild card searches available. Placing a tilde ( ~ ) in front of a search term will search for synonyms of the term but will also search for alternative endings for the term. An example can be found with the search ~eng which returns results for eng, English, engineer and engineering.
The combination of wild cards and synonyms is very useful when searching. A nice tip for working out which synonyms Google will search is to use a string along the lines of "~search -search". The bold terms in your results are the ones being referenced; in this case Finder, Listings, and database along with searcher, searchable, searching, etc.

Brackets within brackets shouldn't cause any problems and can be used as long as they are nested correctly. For example:
("java (code | developer)" | dot.net.nuke) login
This will search for pages that include the phrase "java code", "java developer" or "dot net nuke" and the word login.
Some fairly dramatic strings can be created in this way;
("hire (a | an) (java | html) (coder | designer)" | "html (designer | coder) (in | from) new york")

Although there is no real proximity search tool you can hack one easily enough using the asterisk symbol as a replacement for a single word;
A search using the terms "java * * * design" will more or less, and more more than less, bring back results where the word "java" is three steps from the word "design".

This can be be put to good use discovering the names of people as in this search:
"for more (information | details) (contact | email) (* * | * * * ) project manager"

Permalink Reply by Irina Shamaeva on January 5, 2009 at 12:02am

Hi Simon,
Very interesting ideas to explore here, and great string examples, thanks!

I (slightly) disagree with your point about wildcards. I think that ~xyz will bring words that can be abbreviated as xyz or are "like" xyz (as an example, eng can be an abbreviation for English and engineer) but not everything starting with xyz. Searching for ~consul will bring out embassy but not consultant or consulting.

An additional note on using *. I, too, have noticed that putting one, two or more symbols * works sort-of like proximity search on Google. What Google does, exactly, is somewhat mysterious since Google help says that * stands for "one or more words". Here's an example.
If you look for "learn * java" you will find all of "learn java", "Learn how to enable Java", "Learn EJB - Enterprise Java" etc. But it's true that if you search for "learn ** java" the distance between the words will become larger. Perhaps we could say that * stands for "very few words, in most cases, one"? What do you think?

Simon Townsend said:

Google does have a wild card feature which is actually more useful than most wild card searches available. Placing a tilde ( ~ ) in front of a search term will search for synonyms of the term but will also search for alternative endings for the term. An example can be found with the search ~eng which returns results for eng, English, engineer and engineering.
The combination of wild cards and synonyms is very useful when searching. A nice tip for working out which synonyms Google will search is to use a string along the lines of "~search -search". The bold terms in your results are the ones being referenced; in this case Finder, Listings, and database along with searcher, searchable, searching, etc.

Brackets within brackets shouldn't cause any problems and can be used as long as they are nested correctly. For example:
("java (code | developer)" | dot.net.nuke) login
This will search for pages that include the phrase "java code", "java developer" or "dot net nuke" and the word login.
Some fairly dramatic strings can be created in this way;
("hire (a | an) (java | html) (coder | designer)" | "html (designer | coder) (in | from) new york")

Although there is no real proximity search tool you can hack one easily enough using the asterisk symbol as a replacement for a single word;
A search using the terms "java * * * design" will more or less, and more more than less, bring back results where the word "java" is three steps from the word "design".

This can be be put to good use discovering the names of people as in this search:
"for more (information | details) (contact | email) (* * | * * * ) project manager"

Permalink Reply by Simon Townsend on January 5, 2009 at 12:25am

You're right that not all of the possible alternate endings are searched with the tilde, and it may be that that the algorithm restricts to words that can be shortened to the abbreviated form.
Using your example the extrapolated terms are consulate, consular, and consulado which certainly shows that it is not bringing back anything to do with consultants. Given that ~consult is bringing back terms including consultant, consulting, etc I assume that Google will break it to the nearest exact word and use that as the basis for extrapolation. Quite clever really.

I agree that the asterisk doesn't always represent a single word and "very few words, in most cases, one" is probably the best description for it. Mind you I find that it brings back the required number of replacements more often than not and is incredibly useful as a result. If I could find a way to remove all the noise from my searches I might be a happy man... then again there have been a number of times I've taken a tangent which has proved to be very useful as a result of something that has caught my eye in the clutter.