Update filtered_mc4.py

filtered_mc4.py

@@ -1,6 +1,5 @@
 import json
 import re
-from re import Pattern
 import gzip
 import json
 
@@ -257,7 +256,7 @@ _N_SHARDS_PER_SPLIT = {
     "zu": {"train": 8, "validation": 1},
 }
 
-MC4_MAX_REJECT_PATTERN_OCCURENCE = 1
+MC4_MAX_REJECT_PATTERN_OCCURENCE = 3
 MC4_FILTER_TARGET_FIELD = "text"
 
 
@@ -269,7 +268,7 @@ class FilteredMc4Config(datasets.BuilderConfig):
         *args,
         languages,
         filter_target_field: str=MC4_FILTER_TARGET_FIELD,
-        reject_pattern: str=r"(?!)",
+        reject_patterns: List[str]=[r"(?!)"],
         max_reject_pattern_occurence: int=MC4_MAX_REJECT_PATTERN_OCCURENCE,
         **kwargs,
     ):
@@ -285,13 +284,14 @@ class FilteredMc4Config(datasets.BuilderConfig):
         )
         self.languages = languages
         self.filter_target_field = filter_target_field
-        self.reject_pattern = re.compile(reject_pattern)
+        self.reject_patterns = [re.compile(_) for _ in reject_patterns]
         self.max_reject_pattern_occurence = max_reject_pattern_occurence
 
     def filter(self, example: dict) -> bool:
-        for count, _ in enumerate(self.reject_pattern.finditer(example[self.filter_target_field])):
-            if count == self.max_reject_pattern_occurence:
-                return False
+        for reject_pattern in self.reject_patterns:
+            for count, _ in enumerate(reject_pattern.finditer(example[self.filter_target_field])):
+                if count == self.max_reject_pattern_occurence:
+                    return False
         return True