Google misspellings stats for "britney spears" | Hacker News Reader

Google misspellings stats for "britney spears"

google.com

55 pointsnilsjuenemann35 comments

It seems really weird that every single one of these spells "spears" correctly.

Is this supposed to be a representative list of misspellings? Or given that it's posted under /jobs, is it data for an interview exercise of something like that? (The page doesn't say.)

As I mentioned below, it's used to illustrate that Google works on cool stuff on a page called "Why Work at Google?".

But it's an interesting question: what kind of puzzle could use this as source data? How about something along the following lines: Construct a simple probabilistic program (or FSM, with probabilistically weighted transitions) which outputs variants in proportion to the misspellings in the file.

The fact they have mistakes as bad as "brtittny" but nothing wrong with "spears" suggests to me the data was generated by a query along the lines of "<something like britney> spears" (i.e. the query specified it had to end in "spears").

I remembered seeing these stats before. They are a few years old, right? Google lists the publishing date for those stats as 1 April 2002. Seems like a subset too.

I really enjoy using Google's spelling correction, but what really would blow my mind, is if I one day can search for:

  ntoymru d[rstd

and get results for "britney spears". It seldomly happens when I calibrate my index fingers over the wrong keys.

> ntoymru d[rstd

>It seldomly happens when I calibrate my index fingers over the wrong keys.

I actually do that on purpose a lot... "unreadable password generator".

The trouble is, I think a lot of password crackers now include dictionary words offset on the keyboard in this way, so someone using a dictionary word disguised like this is still at risk of having their password inferred from a hash.

*@97oeh[5@9446Qg975%yq5-J6)qww294ewq43o9ht.

That's a substitution for: I wouldn't worry about that-my passwords are long.

For banking passwords, I use a substitution algorithm that's mathematically based.

Now, when I first saw this, I hypothesized that the number of mistakes would go down roughly with edit distance. Not exactly true, see for yourself:

https://docs.google.com/spreadsheet/ccc?key=0Ao73DTH98IRgdC1...

I used a perl module (http://search.cpan.org/dist/Text-Levenshtein/Levenshtein.pm) to calculate the edit distances.

Many of the differences aren't typoes (classified by edit distance) but sound-alikes.

theoretically one could account for this as edit distance in phoneme space...

Updated the document with a column giving the Soundex code (http://perldoc.perl.org/Text/Soundex.html) for just the first name (since last names are all the same. For the 593 entries, there are only 296 unique Soundex codes.

EDIT: BTW, the original form of the name is Brittany, so the top misspelling technically is not. This is one of the cases, like babe converting to baby where the misspelled and/or affectionate form became dominant.

Know off-hand why "britnay" and "briteny" come up as edit distance 6? Compare "britanny" at 2.

Not a single person misspelled "spears" as "speers" ?

By the time you spelt britney wrong, I would think Google will complete the whole name.

I think they made this list before auto-suggest existed.

Do they only have those stats for "britney spears"? (And if so...why?)

I looked (admittedly not very hard) to see if this was some kind of google lab where you query with some phrase and see the misspellings, but to no avail. That would be interesting - anyone know links?

It's used as an illustrative example on the "Why Work at Google?" page: http://www.google.com/intl/en/jobs/swe/whygoogle/

I thought at first it might be the source data for some kind of programming puzzle, but searching "link:www.google.com/jobs/britney.html site:google.com" did not immediately reveal it. Upon reflection, a puzzle of that sort seems a bit uncharacteristic.

It's worth noting that the Wayback Machine shows the Britney misspelling page existing almost unchanged since March 2002 (http://wayback.archive.org/web/*/http://www.google.com/jobs/...). The back link has varied a bit over the years.

That's interesting, but maybe would have been better if they'd discounted the phrases where the user clicked "Show results for [mis-spelled query] instead"

Some of those look like legitimate queries..

1807 briney spears

Probably just searching for a damn pickle.

Yeah! I agree I think its a bit bold to say that Google 'corrected' "brandi spears" to "britney spears" - there almost certainly are numerous "brandi spears"...

Haha yeah, exactly what I thought.

Actually, it seems these are only the spelling stats for misspelling 'britney' - every single 'spears' is spelt correctly. I was expecting to at least see 'britney speares' listed.

I've always wondered if spelling could be improved through electrical shocks?

Don't overreact! Speaking for myself, I often horrify my sister when I Google for something with a terrible misspelling because sometimes I am a lazy typist; and Google fixes it for me like magic! If only real-life was as easy ;)

I do that same. When it matters I spell correctly (including in IMs). But for google searches or when I have spell check I don't bother.

I often mistype stuff because my fingers are offset on on the keyboard! I try to type "hello world" but it comes out as "y3oo9 294ie" or similar.

I always wondered how often folks do this and why Google doesn't detect it and suggest "did you mean"?

Probably because that adds 6x more possibilities to the search space (and that's assuming consistent transposition in a certain direction).

Isn't the name ' britney' itself a misspelling?

Some background from WP:

"Brittany is a female given name of French and Latin origins, after Brittany, a region of France...

Brittany is also derived from 'Britannia', a 2nd century Roman goddess."

what was the job link to that?

Really? - 2 grittney spears

g is above b on my keygoard.

damn, I was hoping to see britney spores.

Oh come on, half those people are typing one handed.

Sorry.

Reply on news.ycombinator.com