Translation from old to new orthography is done with the script mhr/src/orthography/oldorthography2norm.xfscript.

After having compiled the Mari transducers, the orthography folder contains the transducer oldorthography2norm.lookup.xfst, which will convert input from old to new orthography. There is also an inverse transducer oldorthography2norm.compose.xfst, which is less useful.

The transducer is put into the analysis pipeline between preprocessing and analysis:

cat ~/oldmhr.txt |preprocess|lookup src/orthography/oldorthography2norm.lookup.xfst |grep '\t'|cut -f2|hfst-lookup src/analyser-gt-desc.hfstol

The parts of the pipeline are explained here:

analysis step	comment
cat ~/oldmhr.txt	take input text
preprocess	change it to one word a line
lookup src/orthography/oldorthography2norm.lookup.xfst	send it to the converter
grep ‘\t’	skip the empty lines
cut -f2	pick the second column (the one with new morphology)
hfst-lookup src/analyser-gt-desc.hfstol	and send it to the ordinary analyser
…	and then process it further on just as for modern text.

With the new, fst-based preprocessing we are moving towards, the orthographic converter should be integrated into the overall fst. This is not treated here.