readme

README.md

@@ -0,0 +1,46 @@
+---
+languages:
+- en
+multilinguality:
+- monolingual
+size_categories:
+- 100K<n<1M
+task_categories:
+- conditional-text-generation
+task_ids:
+- summarization
+---
+
+# PubMed dataset for summarization
+
+
+Adapted from this [repo](https://github.com/armancohan/long-summarization).
+Note that original data are pre-tokenized. This dataset returns ' '.join(text).
+This dataset is compatible with the `run_summarization.py` [script](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization) from Transformers if you add this line to the `summarization_name_mapping` variable:
+```python
+"ccdv/pubmed-summarization": ("article", "abstract")
+```
+
+# Cite original article
+```
+@inproceedings{cohan-etal-2018-discourse,
+  title = "A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents",
+  author = "Cohan, Arman  and
+    Dernoncourt, Franck  and
+    Kim, Doo Soon  and
+    Bui, Trung  and
+    Kim, Seokhwan  and
+    Chang, Walter  and
+    Goharian, Nazli",
+  booktitle = "Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)",
+  month = jun,
+  year = "2018",
+  address = "New Orleans, Louisiana",
+  publisher = "Association for Computational Linguistics",
+  url = "https://aclanthology.org/N18-2097",
+  doi = "10.18653/v1/N18-2097",
+  pages = "615--621",
+  abstract = "Neural abstractive summarization models have led to promising results in summarizing relatively short documents. We propose the first model for abstractive summarization of single, longer-form documents (e.g., research papers). Our approach consists of a new hierarchical encoder that models the discourse structure of a document, and an attentive discourse-aware decoder to generate the summary. Empirical results on two large-scale datasets of scientific papers show that our model significantly outperforms state-of-the-art models.",
+}
+```
+

pubmed-summarization.py

@@ -51,7 +51,6 @@ class PubMedSummarizationConfig(datasets.BuilderConfig):
 class PubMedSummarizationDataset(datasets.GeneratorBasedBuilder):
     """PubMedSummarization Dataset."""
     
-    _DOWNLOAD_URL = "https://huggingface.co/datasets/ccdv/pubmed-summarization/resolve/main/"
     _TRAIN_FILE = "train.zip"
     _VAL_FILE = "val.zip"
     _TEST_FILE = "test.zip"
@@ -83,25 +82,10 @@ class PubMedSummarizationDataset(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        """
-        train_path = dl_manager.download_and_extract(self._DOWNLOAD_URL + self._TRAIN_FILE)
-        val_path = dl_manager.download_and_extract(self._DOWNLOAD_URL + self._VAL_FILE)
-        test_path = dl_manager.download_and_extract(self._DOWNLOAD_URL + self._TEST_FILE)
-
-        
-        dl_paths = dl_manager.download_and_extract(self._TRAIN_FILE)
-        train_files = _subset_filenames(dl_paths, datasets.Split.TRAIN)
 
-        dl_paths = dl_manager.download_and_extract(self._VAL_FILE)
-        train_files = _subset_filenames(dl_paths, datasets.Split.TRAIN)
-
-        dl_paths = dl_manager.download_and_extract(self._TEST_FILE)
-        train_files = _subset_filenames(dl_paths, datasets.Split.TRAIN)
-        """
         train_path = dl_manager.download_and_extract(self._TRAIN_FILE) + "/train.txt"
         val_path = dl_manager.download_and_extract(self._VAL_FILE) + "/val.txt"
         test_path = dl_manager.download_and_extract(self._TEST_FILE) + "/test.txt"
-        print("PATHS\n", train_path, val_path, test_path)
         
         return [
             datasets.SplitGenerator(