forbid spaces in token and - in labels

nerval/parse.py

@@ -5,7 +5,7 @@ from pathlib import Path
 NOT_ENTITY_TAG = "O"
 BEGINNING_POS = ["B", "S", "U"]
 
-REGEX_IOB_LINE = re.compile(r"^(.*) ((?:[BIESLU]-|O).*)$")
+REGEX_IOB_LINE = re.compile(r"^(\S*) ((?:[BIESLU]-|O).*)$")
 REGEX_LABEL = re.compile(r"[BIESLU]-(.*)$")
 
 
@@ -45,7 +45,11 @@ def parse_line(index: int, line: str, path: Path):
 
         assert match_iob
 
-        return match_iob.group(1, 2)
+        word, label = match_iob.group(1, 2)
+        # We should have either one - (BLIU-) or none at all (O)
+        assert label.count("-") <= 1
+
+        return word, label
     except AssertionError:
         raise (
             Exception(

tests/test_parse_bio.py

@@ -208,10 +208,7 @@ def test_parse_line(line, word, label):
 
 @pytest.mark.parametrize(
     "line",
-    (
-        ("HiB-ORG"),
-        ("HiB-ORG or maybe not"),
-    ),
+    (("HiB-ORG"), ("HiB-ORG or maybe not"), ("Hello B-surname and L-ocation")),
 )
 def test_parse_line_crash(line):
     with pytest.raises(Exception):