This is a shor desription of how to build a TMX file for CAT Autsch using the gt_sd parallel corpus.
in freecorpus

1. get the latest version of the free corpus:
svn up 

2. extract parallel files for a certain language pair:
java net.sf.saxon.Transform -it main extract_parallel_corpus.xsl lang1=nob lang2=sme inDir=converted

3. index sentences 
corpus-analyze.pl --all --output="./paracosent/$current_lang/$base_n.sent.xml" --only_add_sentences --lang=$current_lang "$file" 

4. align sentences (in gtsvn/tools/alignment-tools/tca2)
perl sentalign_parallel-corpus.pl --dir=DIR --lang1=LANG1 --lang2=LANG2 --anchor=ANCHOR_FILE
Ex.:
perl sentalign_parallel-corpus.pl --dir=0_sme2nob_test --lang1=sme --lang2=nob --anchor=anchor_files/anchor-smenno.txt

5. build the TMX file (see gtsvn/mt/autshumato-ite/conv_test):
 5.1
   perl sentaligned2tmx.pl --aligned=result_tca2 --lang1=nob --lang2=sme
 5.2 put all output into a single TMX file
   java -Xmx2048m net.sf.saxon.Transform -it main unifiyTMX.xsl inDir=tmx-out

============================================