GiellaLT provides an infrastructure for rule-based language technology aimed at minority and indigenous languages, and streamlines building anything from keyboards to speech technology. Read more about Why. See also our Privacy document.
Test data: we are marking up gold-standard documents with a
relatively simple, plain-text
error markup. There are separate
documents for each of the languages we have gold-standard documents
for:
The test data is converted to xml and is stored in
$GTFREE/stable/goldstandard/converted/ when proofread and checked,
before that in $GTFREE/prestable/goldstandard/converted/
You run a speller test using the following command:
new infra:
cd $GTBIG/prooftesting && ./autogen.sh && ./configure && make -
to run all tests for all languages, cd into one of the language
subdirs and
./autogen.sh && ./configure && make to
run the tests for just one language. After ./configure, you
can even cd into one of the speller tool dirs, and run make
there, to run the tests for one speller only.
old infra:
cd $GTHOME/gt && make GTLANG=sma TESTTOOL=pl correct-test