Hi!
I would like to ask if anyone has worked on anything like a reverse bidi
algorithm.
Being a native bidi writer myself, I know that a visually laid out text
may have many logical equivalents; but there are only a few that make
sense and relate to the order the reader may read the text. And I
also agree that this is really locale dependant, (e.g., U+003B SEMICOLON
is considered a neutral punctuation mark in Hebrew, but a strict
left-to-right character in Persian).
But I think that a common framework should exist, for ideas, some
algorithm may assign "reverse bidi" types based on the locale, compute
embedding levels, use that to find the logical ordering of characters, and
then insert implicit or explicit bidi marks for that ordering.
Applications include converting from legacy visual encodings (yes, I know
about the LRO-PDF hack, but I'm talking about things that make sense), and
optical character recognition.
--roozbeh
This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:21:17 EDT