TC A2_readme

TRANSLATION CORPUS ALIGNER (TCA) 2

Home page: http://gandalf.aksis.uib.no/tca2/ (only in Norwegian at the moment)

(See also the file list --files.readme.txt below)

The program is written in Java by Øystein Reigem, AKSIS/UNIFOB, University of Bergen

The text files have to be in marked up in XML and well-formed, but they don't have to have a DTD. The texts have to be divided into sentences (or other unit to be aligned) and each sentence has to have an id. See the sample files RS61E.xml and RS61N.xml (ready for alignment).

Two small support programs are included in the directory. One program divides a text into senteces. This use punctuation and upper case letters to decide where to split. At the moment it does not recognize upper case letters in utf-8. The program will also divide sentences after the two characters ##. The second program is a program to number the <s> and <head> elements. See job-pc.bat/job-linux.sh how to run these programs.

The texts can be encoded in iso-8859-x or utf-8 (this is given in the encoding attribute on the first line of the text), but the anchor list have to be in utf8.

Format of the anchor list:

  • A list of words separated with comma in language 1 / language 2
  • The words can be truncated to the left and right with a star

Sample: 

begin*, began, begun, start* / begyn*, start*

When files are opened in TCA2 then language 1 file has to be open to the left (as file 1).

The program will align <s> and <head> but this can be changed in the Settings panel where you can add other elements like <item> for list elements (with ancestor <list>).

The program is run by opening a command window and executing the alignment.bat file (the alignment.jar program can also be executed directly by clicking, but at the moment you will not see any error messages. They are sent to standard output and are only visible if you use the command window).

Open File 1, File 2, Anchor Words (and click use this anchor list). Check also the settings panel.

The program require Java 1.5.

File Description
Om_aligmentprogrammet.doc Description of program (in Norwegian)
TCA2_demo_docs_20050706.doc some screen dumps of the program (not 100% up to date) with comments
alignment.bat bat file to run in a command window (you will then see error messages)
alignment.jar the alignment program, require Java 1.5
anchor-eng-nor.txt an English - Norwegian anchor list (in UTF-8)
RS61E-1.xml An English text with paragraph mark-up
RS61E-2.xml An English text with paragraph/sentence mark-up
RS61E.xml An English text with sentence id ready for alignment
RS61N-1.xml A Norwegian text with paragraph mark-up
RS61N-2.xml A Norwegian text with paragraph/sentence mark-up
RS61N.xml A Norwegian text with sentence id ready for alignment
-cor.xml, -new.txt Output formats
RS61-aligned.htm Aligned sample files as table in HTML
gen-id-linux program to generate id (run from command line)
gen-id.exe -
job-pc.bat job on pc to run sentence/gen-id
job-linux.sh job on linux
new2htm.exe Merge two -new files to a HTML-table (run from command line)
sentence-linux program to divide text into sentences (command line, not UTF-8 ready)
sentence.exe -