This is the preliminary repo for the Nakoda FST. This project is developed by Atticus G. Harrigan, Antti Arppe, and Jordan Lachler. Development has recieved support from the Univeristy of Alberta's Alberta Language Technology Laboratory, Supporting Indigenous Language Revitalization and the Canadian Indigenous Languages and Literacy Development Institute.
Eventually this repo will be migtrated into the Giellatekno Infrastructure. THIS IS CURRENTLY THE ACTIVE DEVELOPMENT REPO.
For some reason, words with take an prefix for 1Pl are not generating or analyzing. Other forms are fine.
This model considers Nakoda as an infixing language. Stems have special characters that are replaced with infixes. This is controled with flag diacritics.
The majority of the develoment files are in the /src directory and all instructions below assume it as the woeking directory.
-Stems are located in the sto-stems.lexc file.
-Post-stem information is located in the sto-suff.lexc file.
-Infixes (including ones that surface word initially) are located in the sto-infixes.lexc.
-Prefixes for words like maní are found in sto-phon.xfscript.
-Morphosyntactic features +1Sg are located in sto-stems.lexc.
-Morpho-phonological rules are found in sto-phon.xfscript.
To compile the FST you should navagat to the /src folder. The most recent version of the FST is available as a foma binary at sto.foma. You can load this into foma if you would like by running load sto.foma within foma. You can also use a python module like fst-lookup.
If you wish to interact with the FST directly, you can run bash compile.sh. This will concatenate the stems and suffixes lexc files, and will then run foma -l sto-master.xfscript Note that doing this will automatically save the stack to a foma binary, so if you have made changes to the source files, running sto-master.xfscript will automatically change the sto.foma`.
If you wish, you can run bash runtests.sh to test the FST. This will run the items in goldstandard.tsv through the FST, generating surface forms for analyses. The results will then be compared against the goldstandard and any discrepencies will be written to discrepencies.txt. The left column is what is expected and the right column is what was observed.