The L2 transducer is a transducer that models probable L2 errors. Similar transducers are planned for Plains Cree and Northern Haida as well.

It is made as follows:

For lexc, the ordinary files are named as for all other languages, adjectives.lexc, etc. The L2 transducer files for adjectives, nouns and verbs in the stems and affixes folders are named affixes/adjectives-L2.lexc etc. They shall be manually copied over from branches/errorfst.

For twolc, the ordinary file is called sme-phon-L1.twolc (and is automatically copied over from the oldinfra file twol-sme.txt, by our ordinary update script. The L2 file is called sme-phon-L2.twolc, and is our source file for the L2 transducer.

Note the not-so-elegant asymmetry: sme-phon-L1.twolc is a file copied from oldinfra, and shall not be edited. sme-phon-L2.twolc is the source file itself, and shall be edited. The derived file sme-phon.twolc is the copy of either -L1 or -L2, and shall not be edited.

To compile, in langs/sme:

./configure --enable-oahpa # to make the L2 transducer
./configure --disable-oahpa # to **not** make the L2 transducer
make

The normal transducers have their normal names, the L2 transducers are called -oahpa.