Fix the role of raw vs processed words

gilda/ner.py

@@ -140,19 +140,16 @@ def annotate(
             for span in sorted(applicable_spans, reverse=True):
                 # We have to reconstruct a text span while adding spaces
                 # where needed
-                txt_span = ''
                 raw_span = ''
-                for w, rw, c in zip(words[idx:idx+span],
-                                    raw_words[idx:idx+span],
+                for rw, c in zip(raw_words[idx:idx+span],
                                     raw_word_coords[idx:idx+span]):
                     # Figure out if we need a space before this word, then
                     # append the word.
                     spaces = ' ' * (c[0] - len(raw_span) -
                                     raw_word_coords[idx][0])
-                    txt_span += spaces + w
                     raw_span += spaces + rw
                 context = text if context_text is None else context_text
-                matches = grounder.ground(txt_span,
+                matches = grounder.ground(raw_span,
                                           context=context,
                                           organisms=organisms,
                                           namespaces=namespaces)

gilda/tests/test_ner.py

@@ -83,9 +83,9 @@ def test_context_test():
     context_text = "Calcium is released from the ER."
     results = gilda.annotate(text, context_text=context_text)
     assert len(results) == 1
-    assert results[1].matches[0].term.get_curie() == "GO:0005783"
-    assert results[1].text == "ER"
-    assert (results[1].start, results[0].end) == (14, 16)
+    assert results[0].matches[0].term.get_curie() == "GO:0005783"
+    assert results[0].text == "ER"
+    assert (results[0].start, results[0].end) == (14, 16)
 
 
 def test_punctuation_comma_in_entity():