Update script

README.md

@@ -37,10 +37,10 @@ dataset_info:
     sequence: string
   - name: sentences
     sequence: string
-  - name: levels
-    sequence: uint8
   - name: titles_mask
     sequence: uint8
+  - name: levels
+    sequence: uint8
   - name: labels
     sequence:
       class_label:
@@ -49,16 +49,16 @@ dataset_info:
           '1': pos
   splits:
   - name: train
-    num_bytes: 5260132718
+    num_bytes: 4754764877
     num_examples: 582160
   - name: validation
-    num_bytes: 658387335
+    num_bytes: 595209014
     num_examples: 72354
   - name: test
-    num_bytes: 672558301
+    num_bytes: 608033007
     num_examples: 73232
   download_size: 1569504207
-  dataset_size: 6591078354
+  dataset_size: 5958006898
 - config_name: titled
   features:
   - name: id

preprocess_util.py

@@ -1,7 +1,7 @@
 import os
 
 
-def _parse_article(text: str, id: str, drop_titles: bool = False, hier_titles: bool = False):
+def _parse_article(text: str, id: str, drop_titles: bool = False, prepend_title_stack: bool = False):
     def non_empty(s):
         return s and not s.isspace()
     
@@ -37,7 +37,7 @@ def _parse_article(text: str, id: str, drop_titles: bool = False, hier_titles: b
         title = section['title']
         sentences = [sent for sent in section['sentences'] if non_empty(sent)]
 
-        if hier_titles:
+        if prepend_title_stack:
             # Remove irrelevant titles history
             while titles and (last_level := titles[-1][0]) >= level:
                 titles.pop()
@@ -89,7 +89,7 @@ def _parse_article(text: str, id: str, drop_titles: bool = False, hier_titles: b
     return out
 
 
-def parse_split_files(split_path: str, drop_titles: bool = False, hier_titles: bool = False):
+def parse_split_files(split_path: str, drop_titles: bool = False, prepend_title_stack: bool = False):
     for root, dirs, files in os.walk(split_path):
         for fname in files:
             fpath = os.path.join(root, fname)
@@ -100,5 +100,5 @@ def parse_split_files(split_path: str, drop_titles: bool = False, hier_titles: b
                 text = raw_text, 
                 id = fname,
                 drop_titles = drop_titles,
-                hier_titles = hier_titles
+                prepend_title_stack = prepend_title_stack
             )

wiki727k.py

@@ -24,6 +24,8 @@ Usage:
 
 import os
 import datasets
+from dataclasses import dataclass
+from typing import Optional
 
 from .preprocess_util import parse_split_files
 
@@ -72,6 +74,21 @@ _LICENSE = "MIT License"
 _URL = "https://www.dropbox.com/sh/k3jh0fjbyr0gw0a/AACKW_gsxUf282QqrfH3yD10a/wiki_727K.tar.bz2?dl=1"
 
 
+# @dataclass
+# class Wiki727kBuilderConfig(datasets.BuilderConfig):
+#     """BuilderConfig for Wiki727k dataset."""
+
+#     drop_titles: Optional[bool] = None
+#     prepend_title_stack: Optional[bool] = None
+
+#     def __post_init__(self):
+#         # if self.name not in ["default"]:
+#         #     raise ValueError(f"Unknown configuration name: {self.name}")    
+#         if self.drop_titles and self.prepend_title_stack:
+#             raise ValueError("Prepend title stack is not compatible with drop titles.")
+#         super(Wiki727kBuilderConfig, self).__post_init__()
+
+
 # TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case
 class Wiki727k(datasets.GeneratorBasedBuilder):
     """Wiki727k dataset formulated as a sentence-level sequence labelling task for text segmentation."""
@@ -84,7 +101,7 @@ class Wiki727k(datasets.GeneratorBasedBuilder):
 
     # If you need to make complex sub-parts in the datasets with configurable options
     # You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
-    # BUILDER_CONFIG_CLASS = MyBuilderConfig
+    # BUILDER_CONFIG_CLASS = Wiki727kBuilderConfig
 
     # You will be able to load one or the other configurations in the following list with
     # data = datasets.load_dataset('name', 'config1')
@@ -92,8 +109,12 @@ class Wiki727k(datasets.GeneratorBasedBuilder):
         datasets.BuilderConfig(name="titled", version=VERSION, description="Article titles are kept alongside regular sentences in `sentences` attribute, but differentiated with positive values (i.e. 1 as opposed to 0) in `titles_mask` attribute. (Default configuration with all attributes)"),
         datasets.BuilderConfig(name="untitled", version=VERSION, description="Article titles are droped, therefore `sentences` attribute consists of only regular sentences, and `titles_mask` attribute is not present. (Alternative configuration ready for Document Segmentation task)"),
     ]
-
     DEFAULT_CONFIG_NAME = "titled"  # It's not mandatory to have a default configuration. Just use one if it make sense.
+    
+    # BUILDER_CONFIGS = [
+    #     datasets.BuilderConfig(name="default", version=VERSION, description="The only configuration of the dataset."),
+    # ]
+    # DEFAULT_CONFIG_NAME = "default"
 
     def _info(self):
         # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
@@ -120,6 +141,7 @@ class Wiki727k(datasets.GeneratorBasedBuilder):
             }
         )
         
+        # print(self.config)
         if self.config.name == "untitled":
             features.pop("titles_mask")
             features.pop("levels")
@@ -170,8 +192,10 @@ class Wiki727k(datasets.GeneratorBasedBuilder):
     def _generate_examples(self, filepaths: list, split: str):            
         for filepath in filepaths:
             for doc in parse_split_files(filepath, 
-                                         drop_titles=(self.config.name == "untitled"), 
-                                         hier_titles=False):
+                                         drop_titles = (self.config.name == "untitled"), 
+                                         prepend_title_stack = False):
+                                        # drop_titles = self.config.drop_titles,
+                                        # prepend_title_stack = self.config.prepend_title_stack):
                 yield doc['id'], doc