Upload 3 files

README.md

@@ -1,3 +1,63 @@
 ---
-license: cc-by-sa-4.0
+dataset_info:
+  features:
+  - name: title
+    dtype: string
+  - name: text
+    dtype: string
+  - name: wiki_id
+    dtype: int32
+  - name: paragraph_id
+    dtype: int32
+  - name: topic_infer
+    dtype: int64
+  - name: prob
+    dtype: float64
+  splits:
+  - name: train
+    num_bytes: 565706758
+    num_examples: 1139464
+  download_size: 0
+  dataset_size: 565706758
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+task_categories:
+- text-generation
+language:
+- en
+tags:
+- medical
+pretty_name: w
+size_categories:
+- 1M<n<10M
+license: cc
 ---
+# MedWiki from ClinicalCorp
+
+This repo generates on-the-fly the filtered version of the `Cohere/wikipedia-22-12` on medical topic articles using `MaartenGr/BERTopic_Wikipedia`.
+
+## Original Dataset
+
+https://huggingface.co/datasets/Cohere/wikipedia-22-12
+
+## Topic modelling
+
+https://huggingface.co/MaartenGr/BERTopic_Wikipedia
+
+Check the `med_topics.csv` in the git repo for more info on which topics where targeted by prompting `GPT3.5-turbo 0613` over word representations of topics. The original topic list can be obtained from the topic model.
+
+# LICENSE
+
+This dataset is under Wikipedia's licensing which is *CC-BY-SA* : https://creativecommons.org/licenses/by-sa/4.0/ .
+
+# Citation
+
+    @article{corbeil2024iryonlp,
+      title={IryoNLP at MEDIQA-CORR 2024: Tackling the Medical Error Detection & Correction Task On the Shoulders of Medical Agents},
+      author={Jean-Philippe Corbeil},
+      journal={arXiv preprint arXiv:2404.15488},
+      year={2024}
+    }

medwiki_dataset.py

@@ -0,0 +1,95 @@
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""MedWiki."""
+
+import os
+
+import datasets
+import pandas as pd
+
+
+_CITATION = """\
+@article{corbeil2024iryonlp,
+  title={IryoNLP at MEDIQA-CORR 2024: Tackling the Medical Error Detection & Correction Task On the Shoulders of Medical Agents},
+  author={Jean-Philippe Corbeil},
+  journal={arXiv preprint arXiv:2404.15488},
+  year={2024}
+}
+"""
+
+_DESCRIPTION = """\
+This is a filtered version of the `Cohere/wikipedia-22-12` on medical topic articles using `MaartenGr/BERTopic_Wikipedia`. Keep note that some articles in the viewer might seem off topic, but usually they are related in some way (e.g. World War I is linked to the Spanish Flu). This is artefacts of some noise in the topic modelling.
+"""
+_HOMEPAGE = ""
+_LICENSE = "CC-BY-SA"
+_URLS = {
+    "first_domain": "Cohere/wikipedia-22-12",
+}
+
+
+class MedWikiDataset(datasets.GeneratorBasedBuilder):
+    """Medical Wikipedia Articles."""
+
+    VERSION = datasets.Version("0.0.1")
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(name="first_domain", version=VERSION, description="This part of my dataset covers a first domain"),
+    ]
+
+    DEFAULT_CONFIG_NAME = "first_domain"
+
+    def _info(self):
+        features = datasets.Features(
+            {
+                "wiki_id": datasets.Value("int32"),
+                "title": datasets.Value("string"),
+                "text": datasets.Value("string"),
+                "paragraph_id": datasets.Value("int32"),
+            }
+        )
+
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        urls = _URLS[self.config.name]
+        dataset = datasets.load_dataset(urls, "en", trust_remote_code=True, cache_dir="/Users/jcorbeil/Documents/datasets/TEMP")
+        df = pd.read_csv(os.path.join(self.cache_dir, "med_topics.csv"))
+        med_wiki_ids = set(df["wiki_id"].values.tolist())
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "dataset": dataset["train"],
+                    "med_wiki_ids": med_wiki_ids,
+                },
+            ),
+        ]
+
+    def _generate_examples(self, dataset, med_wiki_ids):
+        count = -1
+        for data in dataset:
+            if data["wiki_id"] in med_wiki_ids:
+                count += 1
+                yield count, {
+                    "wiki_id": data["wiki_id"],
+                    "title": data["title"],
+                    "text": data["text"],
+                    "paragraph_id": data["paragraph_id"],
+                }