Greetings all:

It occurs to me now that I might have checked for existing work on the 
lists before I did this, but anyway -- we are in the finishing stages of 
creating scripts that will automatically convert a library's existing 
Romanized MARC Hebrew fields (e.g. "Sefer {dotb}Hatan Torah") into 
Hebrew-script, and add them to the records already in the ILS.  It's 
quite accurate; not bulletproof, but at least it's a way to quickly get 
Hebrew script into thousands of Roman-only records, where many Hebrew 
users (including staff) may not understand the transliteration rules 100%.

The Hebrew conversion itself is done by a PHP script (haven't finished 
learning Perl) acting on a MARC dump of Roman-only Hebrew records in MRK 
(broken MARCedit) format.  This outputs two files of converted fields: 
an XML file for proofing, and a tab-delimited text file for the 
inputting script to devour.  This inputting is done by an Expect script 
using the character-based ILS client.

We are an III shop.  This could presumably be adapted easily enough for 
another ILS, whether using Expect or direct manipulation of database 
tables.  (I'm not volunteering, though...) It would probably be easy 
enough to adapt to another language also, assuming that language were at 
least as predictable in MARC as Hebrew.  (It's pretty good - my list of 
"manual override" words that the auto-algorithm botches is now totaling 
about 35 in preliminary testing.)

Note that I can't imagine automating the other direction, Hebrew- to 
Roman-script, unless there's some algorithm for this already floating 
around out there.

If anyone's interested, I'll clean up the code and open-source it.

Cheers, Shabbat shalom,

Yitzchak Schaffer
Systems Librarian
Touro College Libraries
33 West 23rd Street
New York, NY 10010
Tel (212) 463-0400 x5230
Fax (212) 627-3197
[log in to unmask]