38

I'd like to register misspellings for my domain name, but I don't want to speculatively register a zillion "just guessing" forms. For example, this tool offers the following options in a Keyword Typo Generator

  • skip letters
  • double letters
  • reverse letters
  • skip spaces
  • missed key
  • inserted key

Which is great, and generates dozens if not hundreds of possible domain names... but it doesn't tell me the most crucial thing: which of these misspellings actually happen the most with real live users on the actual Internet?

How can I find actual misspelling data from real users?

|improve this question
25

There are a few resources on Wikipedia, but you'd have to be quite lucky indeed to have a domain that's a plain vanilla English word, common enough to appear here.

You could use the keyword typo generator tool, or something like it, then laboriously enter each one into Google -- which ironically will try to fix your spelling -- then use the "search instead for {misspelled version}" and see how many search results you get for that specific misspelling:

Showing results for treatise

Search instead for tretise

That provides about 117,000 results for "tretise" vs. 17,800,000 results for "treatise". So you now have some idea how common that misspelling is on the whole of the Internet! So you can favor the misspellings that occur more often in the real world.

Perhaps most importantly, I also found these pages:

http://how-to-spell.net/treatise

How to spell treatise?

Correct: treatise.

Common misspellings: tretise - 100%

Which apparently does use actual data!

We have collected misspelling percentages from nearly 15,423,252 spell check sessions on spellchecker.net website (Jan 2010 - Jun 2012).

Ideally I'd love it if Google shared data on misspellings, since I suspect they have a ton more data.. but I'm not sure if they're willing to share, maybe they view automatic spelling correction of mistyped search terms as a competitive advantage.

Would be awesome though if they did!

|improve this answer
  • 2
    Don't forget that you can also ask your users! – Alex L Feb 1 '13 at 9:23
  • 2
    The Google Triliion word corpus surely included both correctly and incorrectly spelled words, so you could use it and lots of word and number crunching to calculate these yourself from raw data Google does share. – hippietrail Feb 1 '13 at 11:07
  • OpenDNS and Google Public DNS definitely have this kind of information, but neither are giving it away but instead using it to improve their own services (OpenDNS has a typo correction feature). You could start your own free/libre DNS service and collect data, I guess. – Christian Davén Feb 1 '13 at 11:19
  • 1
    I don't think using Google results is a good idea (even if impractical). As xkcd points out on his blag, "The “number of results” count that Google gives when you search is clearly fabricated". – rodrigoq Feb 1 '13 at 22:31
9

Interesting problem. You could use google's keyword search tool at https://adwords.google.com/o/KeywordTool to find out how many searches per month are performed for each of your domain name typos (you'd have to generate a typo list with the tool you mention above). Not perfect, but would represent actual user typos and give you some decent data to go on. (I just tested it and got some results that seemed reasonable.)

|improve this answer
  • Sadly, this is no longer true, now that we have Hummingbird. Google Keyword Tool has been deprecated. It was replaced with Keyword Planner, a Google AdSense product. – Ellie Kesselman Nov 12 '13 at 1:22
4

Interesting question, analyzing the search results as described here should definately be a good starting position but is sensitive to false positives caused by other websites with names similar to yours. It may also miss common errors that you did not think about, looking at the list I would say analysing the switching of words may be interesting as well.

A method that is particularly good at discovering which errors are most commonly made by common visitors is by asking them to type in the name. Of course not all websites are suitable for this, assuming the page is interactive it should make for a nice fit.

Rather than using regular captcha's just show a few of your site logos with (part of) the name written there and ask people to type it in before they can do their thing.

The data is not going to be based on a huge population like google search, but it will be exactly the target group and the results should therefore be quite powerfull.

|improve this answer
1

You could try this domain typo finder. Search options the following, individually or in combination:

  • QWERTY keyboard slips
  • Letter swap
  • Sticky keys, for extra or missing letters
  • Look-a-like's such as the letter l versus the number 1

There are three "viewpoints" for searching. Registrant is the default. DNS is a sortable view showing the name server and IP address of each typo. TLD finds typos in names in any of these six TLDs: .com .net .org .biz .us .info and if it is registered.

Is it any different than OP's keyword tool? Maybe. It claims to "find common typo's associated with domain names". That means the data was collected from users as they manually typed domain names into their browser navigation bar.

  • "Sticky keys" search includes common variants, with or without hyphens. Those kind of typos aren't due to physically gummed-up, sticky keys, rather, errors in human perception. That indicates use of something better than a rule based, expert-system approach.
|improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.