Test that the Latin ITTB lemmatizer is marked as caseless. Check that…

… the results for capitalized text is as expected

stanza/tests/pipeline/test_lemmatizer.py

@@ -120,3 +120,17 @@ def test_caseless_lemmatizer():
     doc = nlp("Jennifer has nice Antennae")
     assert doc.sentences[0].words[-1].lemma == 'antenna'
 
+def test_latin_caseless_lemmatizer():
+    """
+    Test the Latin caseless lemmatizer
+    """
+    nlp = stanza.Pipeline('la', package='ittb', processors='tokenize,pos,lemma', model_dir=TEST_MODELS_DIR, download_method=None)
+    lemmatizer = nlp.processors['lemma']
+    assert lemmatizer.config['caseless']
+
+    doc = nlp("Quod Erat Demonstrandum")
+    expected_lemmas = "qui sum demonstro".split()
+    assert len(doc.sentences) == 1
+    assert len(doc.sentences[0].words) == 3
+    for word, expected in zip(doc.sentences[0].words, expected_lemmas):
+        assert word.lemma == expected

stanza/tests/setup.py

@@ -40,6 +40,8 @@
 stanza.download(lang='en', model_dir=models_dir, logging_level='info')
 stanza.download(lang="en", model_dir=models_dir, package=None, processors={"ner":"ncbi_disease"})
 stanza.download(lang='fr', model_dir=models_dir, logging_level='info')
+# Latin ITTB has no case information for the lemmatizer
+stanza.download(lang='la', model_dir=models_dir, package='ittb', logging_level='info')
 stanza.download(lang='zh', model_dir=models_dir, logging_level='info')
 # useful not just for verifying RtL, but because the default Arabic has a unique style of xpos tags
 stanza.download(lang='ar', model_dir=models_dir, logging_level='info')