add data and loader

aspectemo.py

@@ -0,0 +1,104 @@
+import csv
+from typing import List, Generator, Tuple, Dict
+
+import datasets
+from datasets import DownloadManager
+from datasets.info import SupervisedKeysData
+
+_DESCRIPTION = """AspectEmo 1.0 dataset: Multi-Domain Corpus of Consumer Reviews for Aspect-Based 
+                Sentiment Analysis"""
+
+_CLASSES = ['O',
+            'B-a_plus_m',
+            'B-a_minus_m',
+            'B-a_zero',
+            'B-a_minus_s',
+            'B-a_plus_s',
+            'B-a_amb',
+            'B-a_minus_m:B-a_minus_m',
+            'B-a_minus_m:B-a_minus_m:B-a_minus_m',
+            'B-a_plus_m:B-a_plus_m',
+            'B-a_plus_m:B-a_plus_m:B-a_plus_m',
+            'B-a_zero:B-a_zero:B-a_zero',
+            'B-a_zero:B-a_zero',
+            'I-a_plus_m',
+            'B-a_zero:B-a_plus_m',
+            'B-a_minus_m:B-a_zero',
+            'B-a_minus_s:B-a_minus_s:B-a_minus_s',
+            'B-a_amb:B-a_amb',
+            'I-a_minus_m',
+            'B-a_minus_s:B-a_minus_s',
+            'B-a_plus_s:B-a_plus_s:B-a_plus_s',
+            'B-a_plus_m:B-a_plus_m:B-a_plus_m:B-a_plus_m:B-a_plus_m:B-a_plus_m',
+            'B-a_plus_m:B-a_amb',
+            'B-a_minus_m:B-a_plus_m',
+            'B-a_amb:B-a_amb:B-a_amb',
+            'I-a_zero',
+            'B-a_plus_s:B-a_plus_s',
+            'B-a_plus_m:B-a_plus_s',
+            'B-a_plus_m:B-a_zero',
+            'B-a_zero:B-a_zero:B-a_zero:B-a_zero:B-a_zero:B-a_zero',
+            'B-a_zero:B-a_minus_m',
+            'B-a_amb:B-a_plus_s',
+            'B-a_zero:B-a_minus_s']
+
+_URLS = {
+    "train": "https://huggingface.co/datasets/clarin-pl/aspectemo/resolve/main/data/train.tsv",
+    "validation": "https://huggingface.co/datasets/clarin-pl/aspectemo/resolve/main/data/val.tsv",
+    "test": "https://huggingface.co/datasets/clarin-pl/aspectemo/resolve/main/data/test.tsv",
+}
+
+
+class AspectEmo(datasets.GeneratorBasedBuilder):
+    def _info(self) -> datasets.DatasetInfo:
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "orth": datasets.Sequence(datasets.Value("string")),
+                    "ctag": datasets.Sequence(datasets.Value("string")),
+                    "sentiment": datasets.Sequence(datasets.features.ClassLabel(
+                        names=_CLASSES,
+                        num_classes=len(_CLASSES)
+                    )),
+                }
+            ),
+            supervised_keys=SupervisedKeysData(input="orth", output="sentiment"),
+            homepage="https://clarin-pl.eu/dspace/handle/11321/849",
+        )
+
+    def _split_generators(self, dl_manager: DownloadManager) -> List[datasets.SplitGenerator]:
+        urls_to_download = _URLS
+        downloaded_files = dl_manager.download_and_extract(urls_to_download)
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"filepath": downloaded_files["train"]},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                gen_kwargs={"filepath": downloaded_files["validation"]},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={"filepath": downloaded_files["test"]},
+            ),
+        ]
+
+    def _generate_examples(
+            self, filepath: str
+    ) -> Generator[Tuple[int, Dict[str, str]], None, None]:
+        with open(filepath, "r", encoding="utf-8") as f:
+            reader = csv.reader(f, delimiter="\t", quoting=csv.QUOTE_NONE)
+            next(reader, None)  # skip header
+            id_, orth, ctag, sentiment = set(), [], [], []
+            for line in reader:
+                if not line:
+                    assert len(id_) == 1
+                    yield id_.pop(), {"orth": orth, "ctag": ctag, "sentiment": sentiment, }
+                    id_, orth, ctag, sentiment = set(), [], [], []
+                else:
+                    id_.add(line[0])
+                    orth.append(line[1])
+                    ctag.append(line[2])
+                    sentiment.append(line[3])