Usage stats?

Phillips, Addison addison at
Fri Mar 27 15:40:35 CDT 2015

What you might be looking for would be the CLDR project’s “exemplar sets” (see for example [1]), which describes which characters are customarily used for a given language and which are sometimes used. However, this is not the same thing as statistical distribution. One of the points of Unicode is that any character can be used at any time in any document—regardless of language.


From: Unicode [mailto:unicode-bounces at] On Behalf Of Michael Norton
Sent: Friday, March 27, 2015 1:25 PM
To: John D. Burger
Cc: Vint Cerf; Unicode at
Subject: Re: Usage stats?

Just using the tools and formulations we have at present ought to allow Unicode to produce a usage set without indexing the entire web which would provide implementors with an indication of variances for traffic, overflow, and override purposes relative to users of the standard.  If the figure varies significantly from page:website, website:region, region:language, for example, it simplifies our ability to standardize the set.

I have particular concerns, but, like Google, they are proprietary.

On Fri, Mar 27, 2015 at 4:23 PM, John D. Burger <john at<mailto:john at>> wrote:
On Mar 27, 2015, at 15:57 , Michael Norton <michaelanortonster at<mailto:michaelanortonster at>> wrote:

Why wouldn't Unicode itself have it?

Because as Ken explained, acquiring (and constantly updating) such statistics would require roughly the effort that Google puts into its crawler. And it wouldn't include all the printed material that isn't on the web.

Turning your question around, why would Unicode have this information? What would be the value, and how would it be worth the (considerable) effort required?

- John Burger

On Fri, Mar 27, 2015 at 1:07 PM, Ken Whistler <kenwhistler at<mailto:kenwhistler at>> wrote:
Search engine companies (and in particular, Google) have such
information squirreled away in their index databases, at least as
far as usage stats for Unicode characters on the web go -- but it
is proprietary information, and they generally don't publish
information about such statistics.

Perhaps there are researchers out there who have set web crawlers
on a mission to generate such web statistics for publication, and maybe
somebody on this list knows of such research -- but it would be
virtually impossible to generate such information for the much
wider collection of documents and data that are not easily accessible
for web indexing. (Behind password walls, in pdf document archives,
in proprietary databases, ... ) As an example of why this is a problem,
consider the fact that there are *peta*bytes of information picked up
and stored in databases from scanners and other devices used at
tens of millions of retail points of sale. Such data, by its nature, would tend
to skew heavily towards use of ASCII a-z and digits 0-9 in its
character data. How would you end up weighting such (mostly
publicly inaccessible) data in trying to count up for overall statistics
on character use?

There are more traditional usage count studies that focus on
counts of character frequency within single language orthographies
in single scripts (e.g., letter frequences for French text), but I don't
think that is what you were asking about.

Here is some discussion of a similar question posted on stackoverflow:


On 3/27/2015 9:31 AM, Michael Norton wrote:
Hello and thank you for an incredible service (just joining the list).   Is there a list of usage statistics per character of the Unicode set available somewhere?

Unicode mailing list
Unicode at<mailto:Unicode at>


Michael A. Norton, B.A. Cinema, M.P.A.
My Cinema Home:<>

"All great actors are mere mathematical masters of speech and the human body."
[Image removed by sender.]

Unicode mailing list
Unicode at<mailto:Unicode at>


Michael A. Norton, B.A. Cinema, M.P.A.
My Cinema Home:

"All great actors are mere mathematical masters of speech and the human body."
[Image removed by sender.]

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 594 bytes
Desc: image001.jpg
URL: <>

More information about the Unicode mailing list