Updated Old Czech tests.

Author: dan-zeman

udapi/block/ud/cs/addmwt.py

@@ -55,6 +55,18 @@
         'shape': 'subtree',
     }
 
+# Old Czech 'toliť' (special case with 3 subtokens; general -ť will be solved dynamically below).
+MWTS['toliť'] = {
+    'form':   'to li ť',
+    'lemma':  'ten li ť',
+    'upos':   'DET SCONJ PART',
+    'xpos':   '* J,------------- TT-------------',
+    'feats':  '* _ _',
+    'deprel': '* mark discourse',
+    'main':   0,
+    'shape':  'siblings'
+}
+
 
 
 class AddMwt(udapi.block.ud.addmwt.AddMwt):

udapi/block/ud/cs/markfeatsbugs.py

@@ -66,15 +66,15 @@ def process_node(self, node):
                     'Animacy': ['Anim', 'Inan'],
                     'Number': ['Sing', 'Dual', 'Plur'],
                     'Case': ['Nom', 'Gen', 'Dat', 'Acc', 'Voc', 'Loc', 'Ins'],
-                    'NameType': ['Giv', 'Sur', 'Geo', 'Nat'],
+                    'NameType': ['Giv', 'Sur', 'Geo', 'Nat', 'Com', 'Pro', 'Oth'],
                     'Foreign': ['Yes'],
                     'Abbr': ['Yes']})
             else:
                 self.check_allowed_features(node, {
                     'Gender': ['Masc', 'Fem', 'Neut'],
                     'Number': ['Sing', 'Dual', 'Plur'],
                     'Case': ['Nom', 'Gen', 'Dat', 'Acc', 'Voc', 'Loc', 'Ins'],
-                    'NameType': ['Giv', 'Sur', 'Geo', 'Nat'],
+                    'NameType': ['Giv', 'Sur', 'Geo', 'Nat', 'Com', 'Pro', 'Oth'],
                     'Foreign': ['Yes'],
                     'Abbr': ['Yes']})
         # ADJECTIVES ###########################################################
@@ -385,7 +385,9 @@ def process_node(self, node):
                 })
             # Relative possessive determiners 'jehož' and 'jejichž' behave similarly
             # to the personal possessive determiners but they do not have Person.
-            elif re.match(r'^(jeho|jejich|j[ií]ch)ž(e|to)?$', node.form.lower()):
+            # Normally determiners do not change j->n after prepositions but we
+            # have an example in Old Czech (štěpové zlatí, na nichžto větviech...)
+            elif re.match(r'^(jeho|jejich|[jn][ií]ch)ž(e|to)?$', node.form.lower()):
                 self.check_required_features(node, ['PronType', 'Poss', 'Number[psor]'])
                 self.check_allowed_features(node, {
                     'PronType': ['Rel'],
@@ -549,13 +551,14 @@ def process_node(self, node):
                 })
             else:
                 if node.feats['NumType'] == 'Sets':
-                    # 'jedny', 'dvoje', 'troje', 'čtvery'
+                    # 'jedny', 'dvoje', 'oboje', 'troje', 'čtvery'
                     # Number should perhaps be only Plur because the counted noun will be Plur.
                     # Gender is not annotated in PDT but there are different forms ('jedni' vs. 'jedny',
                     # and in Old Czech also 'dvoji' vs. 'dvoje'), so we should allow Gender (and Animacy).
                     self.check_required_features(node, ['NumType', 'NumForm', 'Number', 'Case'])
                     self.check_allowed_features(node, {
                         'NumType': ['Sets'],
+                        'PronType': ['Tot'], # for 'oboje'
                         'NumForm': ['Word'],
                         'Gender': ['Masc', 'Fem', 'Neut'],
                         'Animacy': ['Anim', 'Inan'],

udapi/block/ud/fixadvmodbyupos.py

@@ -69,6 +69,8 @@ def process_node(self, node):
                 node.deprel = 'nmod'
             elif node.upos == 'ADJ':
                 node.deprel = 'amod'
+            elif node.upos == 'NUM':
+                node.deprel = 'nummod'
             elif node.upos == 'ADV':
                 node.deprel = 'advmod'
             elif node.upos == 'AUX':
@@ -77,13 +79,21 @@ def process_node(self, node):
                 node.deprel = 'dep'
             elif node.upos == 'SCONJ':
                 node.deprel = 'mark'
+            elif node.upos == 'CCONJ':
+                node.deprel = 'cc'
             elif node.upos == 'X':
                 node.deprel = 'dep'
         elif node.udeprel == 'nummod':
-            if node.upos == 'PRON':
+            if node.upos == 'ADJ':
+                node.deprel = 'amod'
+            elif node.upos == 'PRON':
                 node.deprel = 'nmod'
             elif node.upos == 'DET':
                 node.deprel = 'det'
+            elif node.upos == 'ADP':
+                node.deprel = 'case'
         elif node.udeprel == 'punct':
             if node.upos != 'PUNCT':
                 node.deprel = 'dep'
+        elif node.udeprel == 'obl' and node.parent.upos in ['NOUN', 'PROPN', 'PRON'] and node.parent.udeprel in ['nsubj', 'obj', 'iobj', 'obl', 'vocative', 'dislocated', 'expl', 'nmod']:
+            node.deprel = 'nmod'