Support streaming

biology_how_why_corpus.py

@@ -21,10 +21,9 @@ used for the question-answering system described in the paper “Discourse Compl
 Answer Reranking” (ACL 2014).
 """
 
-import os
 import xml.dom.minidom as xml
-from itertools import chain, count
-from typing import Dict, List, Tuple
+from itertools import count
+from typing import BinaryIO, Dict, List, Tuple
 
 import datasets
 
@@ -133,53 +132,46 @@ class BiologyHowWhyCorpusDataset(datasets.GeneratorBasedBuilder):
 
     def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
         urls = _URLS[_DATASETNAME]
-        data_dir = dl_manager.download_and_extract(urls)
+        archive_path = dl_manager.download(urls)
 
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "how_path": os.path.join(
-                        data_dir, "BiologyHowWhyCorpus", "GoldStandardVulcanHOW.all.xml"
-                    ),
-                    "why_path": os.path.join(
-                        data_dir, "BiologyHowWhyCorpus", "GoldStandardVulcanWHY.all.xml"
-                    ),
+                    "archive": dl_manager.iter_archive(archive_path),
                 },
             ),
         ]
 
-    def _generate_examples(self, how_path: str, why_path: str) -> Tuple[int, Dict]:
+    def _generate_examples(self, archive: Tuple[str, BinaryIO]) -> Tuple[int, Dict]:
 
         uid = count(0)
 
         if self.config.schema == "source":
-            for question in chain(
-                self._parse_questions(how_path, "how"),
-                self._parse_questions(why_path, "why"),
-            ):
-                yield next(uid), question
+            for path, file in archive:
+                question_type = path.split(".")[-3][-3:].lower()
+                for question in self._parse_questions(file, question_type):
+                    yield next(uid), question
 
         elif self.config.schema == "bigbio_qa":
-            for question in chain(
-                self._parse_questions(how_path, "how"),
-                self._parse_questions(why_path, "why"),
-            ):
-                for answer in question["answers"]:
-                    id = next(uid)
-                    yield id, {
-                        "id": id,
-                        "question_id": next(uid),
-                        "document_id": answer["docid"],
-                        "question": question["text"],
-                        "type": question["type"],
-                        "choices": [],
-                        "context": "",
-                        "answer": [answer["justification"]],
-                    }
+            for path, file in archive:
+                question_type = path.split(".")[-3][-3:].lower()
+                for question in self._parse_questions(file, question_type):
+                    for answer in question["answers"]:
+                        guid = next(uid)
+                        yield guid, {
+                            "id": guid,
+                            "question_id": next(uid),
+                            "document_id": answer["docid"],
+                            "question": question["text"],
+                            "type": question["type"],
+                            "choices": [],
+                            "context": "",
+                            "answer": [answer["justification"]],
+                        }
 
-    def _parse_questions(self, path: str, type: str):
-        collection = xml.parse(path).documentElement
+    def _parse_questions(self, path_or_file: BinaryIO, question_type: str):
+        collection = xml.parse(path_or_file).documentElement
         questions = collection.getElementsByTagName("question")
         for question in questions:
             text = question.getElementsByTagName("text")[0].childNodes[0].data
@@ -202,4 +194,4 @@ class BiologyHowWhyCorpusDataset(datasets.GeneratorBasedBuilder):
                         "sentences": sentences,
                     }
                 )
-            yield {"text": text, "type": type, "answers": answers_}
+            yield {"text": text, "type": question_type, "answers": answers_}