RE: Character set conversion question

From: Dreiheller, Albrecht (
Date: Wed Jun 17 2009 - 02:46:00 CDT

  • Next message: Jeroen Ruigrok van der Werven: "Re: Jyutping Phrase Box to be removed (was: Unihan database: kCangjie field)"

    The brute force approach should be table driven, using a data model
    capable of composite conversions and "reverse misinterpretations" as Björn Höhrmann described it.
    The first few table entries may be manually created, containing favorites.
    After that, one-stage generic entries follow, using all known encodings.
    After that, two-stage generic entries follow, combining two one-stage conversions.
    And so on.
    The user will then choose the applicable decoding by checking the results visually in
    an interactive user interface. (MS Word has such a dialog when opening .txt files with unknown encoding).
    The user's choice may then be used to set up a ranking within the table to find the
    most common encodings (or "misencodings") first, like popular web search engines do it.


    -----Original Message-----
    From: [] On Behalf Of Leo Broukhis
    Sent: Wednesday, June 17, 2009 7:03 AM
    To: Bjoern Hoehrmann
    Cc: unicode Unicode Discussion
    Subject: Re: Character set conversion question

    That's exactly my question: how to organize the brute force approach
    with 217 character maps in /usr/share/i18n/charmaps/ ?


    On Tue, Jun 16, 2009 at 3:38 PM, Bjoern Hoehrmann<> wrote:
    > * Leo Broukhis wrote:
    >>What would be a way to find out what character set conversions were
    >>applied to the text?
    > Where the brute force approach fails and you have not misanalyzed the
    > byte stream (copy and paste from a mail program may be unreliable) it
    > is likely that you either have not tried enough encodings, or the en-
    > coding is the result of function composition, for example, it might
    > have been ISO-8859-X which is then interpreted as ISO-8859-Y and then
    > encoded using ISO-8859-Z by some process; a popular example is UTF-8
    > encoded data re-interpreted as ISO-8859-1 and re-encoded as UTF-8.
    > Then your brute force search has to include such compositions aswell.
    > --
    > Björn Höhrmann · ·
    > Am Badedeich 7 · Telefon: +49(0)160/4415681 ·
    > 25899 Dagebüll · PGP Pub. KeyID: 0xA4357E78 ·

    This archive was generated by hypermail 2.1.5 : Wed Jun 17 2009 - 02:49:19 CDT