Fix scorer bug for NER, related to ambiguity between missing annotations and misaligned tokens

spacy/gold.pyx

@@ -272,8 +272,8 @@ cdef class GoldParse:
         self.words = [None] * len(doc)
         self.tags = [None] * len(doc)
         self.heads = [None] * len(doc)
-        self.labels = [''] * len(doc)
-        self.ner = ['-'] * len(doc)
+        self.labels = [None] * len(doc)
+        self.ner = [None] * len(doc)
 
         self.cand_to_gold = align([t.orth_ for t in doc], words)
         self.gold_to_cand = align(words, [t.orth_ for t in doc])

spacy/scorer.py

@@ -87,7 +87,7 @@ class Scorer(object):
         gold_ents = set(tags_to_entities([annot[-1] for annot in gold.orig_annot]))
         for id_, word, tag, head, dep, ner in gold.orig_annot:
             gold_tags.add((id_, tag))
-            if dep is not None and dep.lower() not in punct_labels:
+            if dep not in (None, "") and dep.lower() not in punct_labels:
                 gold_deps.add((id_, head, dep.lower()))
         cand_deps = set()
         cand_tags = set()

spacy/syntax/ner.pyx

@@ -106,7 +106,7 @@ cdef class BiluoPushDown(TransitionSystem):
                 self.freqs[ENT_TYPE][0] += 1
 
     cdef Transition lookup_transition(self, object name) except *:
-        if name == '-':
+        if name == '-' or name == None:
             move_str = 'M'
             label = 0
         elif '-' in name: