Add a new augmentation method with Textattack: random character inser…

…tion

dataset/preprocessing/data_processor.py

@@ -17,6 +17,7 @@
 from enum import Enum
 
 from textattack.augmentation import Augmenter
+from textattack.transformations import WordSwapRandomCharacterInsertion
 from textattack.transformations.word_swaps.word_swap_neighboring_character_swap import WordSwapNeighboringCharacterSwap
 
 class MythLabels(Enum):
@@ -75,11 +76,19 @@ def _load_data(self):
     class DataAugmentator:
         def __init__(self, outer_instance):
             self.data_processor = outer_instance
-            self.augmenter = Augmenter(
+
+            self.swap_augmenter = Augmenter(
                 transformation=WordSwapNeighboringCharacterSwap(),
+                pct_words_to_swap=0.4,
+                transformations_per_example=random.randint(1, 3)
+            )
+
+            self.augmenter_insert = Augmenter(
+                transformation=WordSwapRandomCharacterInsertion(),
                 pct_words_to_swap=0.5,
-                transformations_per_example=3
+                transformations_per_example=random.randint(1, 3)
             )
+
         def augment(self):
             """
 
@@ -115,8 +124,10 @@ def augment(self):
                         augmented_data.append(f"{label} {name_no_spaces}")
 
                     # Esto es de textattack, creo que será buena idea... veamos.
-                    augmented_names = self.augmenter.augment(name)
-                    for aug_name in augmented_names:
+                    swapped_names = self.swap_augmenter.augment(name)
+                    insertions_names = self.augmenter_insert.augment(name)
+
+                    for aug_name in swapped_names + insertions_names:
                         augmented_data.append(f"{label} {aug_name}")
 
             return augmented_data