Update jparacrawl.py

jparacrawl.py

@@ -158,7 +158,7 @@ class JParaCrawl(datasets.GeneratorBasedBuilder):
                 break
         assert df is not None, extracted_path
 
-        def _split(line: str, col: int):
+        def _split(line: str, col: int) -> str:
             return line.split('\t', 4)[col].strip()
 
         df['domain'] = df[0].apply(partial(_split, col=0))
@@ -168,7 +168,7 @@ class JParaCrawl(datasets.GeneratorBasedBuilder):
         df['ja'] = df[0].apply(partial(_split, col=4))
         df = df.drop_duplicates(subset=[non_ja, 'ja'])
 
-        def _normalize(s: str):
+        def _normalize(s: str) -> str:
             return unicodedata.normalize("NFKC", s).replace('\t', ' ').strip()
 
         _id = 0
@@ -178,8 +178,8 @@ class JParaCrawl(datasets.GeneratorBasedBuilder):
                 "url": row["url"],
                 "probability": float(row["probability"]),
                 "translation": {
-                    non_ja: _normalize(row[non_ja]: str),
-                    "ja": _normalize(row["ja"]: str),
+                    non_ja: _normalize(row[non_ja]),
+                    "ja": _normalize(row["ja"]),
                 },
             }
             # Make sure that both translations are non-empty.