COVID-Fact entailment.

README.md

@@ -1,15 +1,14 @@
 ---
 annotations_creators:
 - expert-generated
-language:
-- en
 language_creators:
 - found
+language:
+- en
 license:
 - cc-by-nc-2.0
 multilinguality:
 - monolingual
-pretty_name: SciFact
 size_categories:
 - 1K<n<10K
 source_datasets:
@@ -18,7 +17,7 @@ task_categories:
 - text-classification
 task_ids:
 - fact-checking
-paperswithcode_id: scifact
+pretty_name: CovidFact
 dataset_info:
   features:
   - name: claim_id
@@ -37,17 +36,17 @@ dataset_info:
     sequence: int32
   splits:
   - name: train
-    num_bytes: 1649655
-    num_examples: 919
-  - name: validation
-    num_bytes: 605262
-    num_examples: 340
-  download_size: 3115079
-  dataset_size: 2254917
+    num_bytes: 1547185
+    num_examples: 940
+  - name: test
+    num_bytes: 523542
+    num_examples: 317
+  download_size: 3610222
+  dataset_size: 2070727
 ---
 
 
-# Dataset Card for "scifact_entailment"
+# Dataset Card for "covidfact_entailment"
 
 ## Table of Contents
 
@@ -59,17 +58,12 @@ dataset_info:
 
 ## Dataset Description
 
-- **Homepage:** [https://scifact.apps.allenai.org/](https://scifact.apps.allenai.org/)
-- **Repository:** <https://github.com/allenai/scifact>
-- **Paper:** [Fact or Fiction: Verifying Scientific Claims](https://aclanthology.org/2020.emnlp-main.609/)
+- **Repository:** <https://github.com/asaakyan/covidfact>
 - **Point of Contact:** [David Wadden](mailto:[email protected])
 
 ### Dataset Summary
 
-SciFact, a dataset of 1.4K expert-written scientific claims paired with evidence-containing abstracts, and annotated with labels and rationales.
-
-For more information on the dataset, see [allenai/scifact](https://huggingface.co/datasets/allenai/scifact).
-This has the same data, but reformatted as an entailment task. A single instance includes a claim paired with a paper title and abstract, together with an entailment label and a list of evidence sentences (if any).
+COVID-FACT is a dataset of claims about COVID-19. For this version of the dataset, we follow the preprocessing from the MultiVerS modeling paper https://github.com/dwadden/multivers, verifying claims against abstracts of scientific research articles. Entailment labels and rationales are included.
 
 ## Dataset Structure
 

covidfact_entailment.py

@@ -7,27 +7,30 @@ import json
 
 
 _CITATION = """\
-@inproceedings{Wadden2020FactOF,
-  title={Fact or Fiction: Verifying Scientific Claims},
-  author={David Wadden and Shanchuan Lin and Kyle Lo and Lucy Lu Wang and Madeleine van Zuylen and Arman Cohan and Hannaneh Hajishirzi},
-  booktitle={EMNLP},
-  year={2020},
+@article{Saakyan2021COVIDFactFE,
+  title={COVID-Fact: Fact Extraction and Verification of Real-World Claims on COVID-19 Pandemic},
+  author={Arkadiy Saakyan and Tuhin Chakrabarty and Smaranda Muresan},
+  journal={ArXiv},
+  year={2021},
+  volume={abs/2106.03794},
+  url={https://api.semanticscholar.org/CorpusID:235364036}
 }
 """
 
+
 _DESCRIPTION = """\
-SciFact, a dataset of 1.4K expert-written scientific claims paired with evidence-containing abstracts, and annotated with labels and rationales.
+COVID-FACT is a dataset of claims about COVID-19. For this version of the dataset, we follow the preprocessing from the MultiVerS modeling paper https://github.com/dwadden/multivers, verifying claims against abstracts of scientific research articles. Entailment labels and rationales are included.
 """
 
-_URL = "https://scifact.s3-us-west-2.amazonaws.com/release/latest/data.tar.gz"
+_URL = "https://scifact.s3.us-west-2.amazonaws.com/longchecker/latest/data.tar.gz"
 
 
 def flatten(xss):
     return [x for xs in xss for x in xs]
 
 
-class ScifactEntailmentConfig(datasets.BuilderConfig):
-    """BuilderConfig for Scifact"""
+class CovidFactEntailmentConfig(datasets.BuilderConfig):
+    """builderconfig for covidfact"""
 
     def __init__(self, **kwargs):
         """
@@ -35,19 +38,19 @@ class ScifactEntailmentConfig(datasets.BuilderConfig):
         Args:
             **kwargs: keyword arguments forwarded to super.
         """
-        super(ScifactEntailmentConfig, self).__init__(
+        super(CovidFactEntailmentConfig, self).__init__(
             version=datasets.Version("1.0.0", ""), **kwargs
         )
 
 
-class ScifactEntailment(datasets.GeneratorBasedBuilder):
-    """TODO(scifact): Short description of my dataset."""
+class CovidFactEntailment(datasets.GeneratorBasedBuilder):
+    """TODO(covidfact): Short description of my dataset."""
 
-    # TODO(scifact): Set up version.
+    # TODO(covidfact): Set up version.
     VERSION = datasets.Version("0.1.0")
 
     def _info(self):
-        # TODO(scifact): Specifies the datasets.DatasetInfo object
+        # TODO(covidfact): Specifies the datasets.DatasetInfo object
 
         features = {
             "claim_id": datasets.Value("int32"),
@@ -92,13 +95,14 @@ class ScifactEntailment(datasets.GeneratorBasedBuilder):
         # download and extract URLs
         archive = dl_manager.download(_URL)
         for path, f in dl_manager.iter_archive(archive):
-            if path == "data/corpus.jsonl":
+            # The claims are too similar to paper titles; don't include.
+            if path == "data/covidfact/corpus_without_titles.jsonl":
                 corpus = self._read_tar_file(f)
                 corpus = {x["doc_id"]: x for x in corpus}
-            elif path == "data/claims_train.jsonl":
+            elif path == "data/covidfact/claims_train.jsonl":
                 claims_train = self._read_tar_file(f)
-            elif path == "data/claims_dev.jsonl":
-                claims_validation = self._read_tar_file(f)
+            elif path == "data/covidfact/claims_test.jsonl":
+                claims_test = self._read_tar_file(f)
 
         return [
             datasets.SplitGenerator(
@@ -111,12 +115,12 @@ class ScifactEntailment(datasets.GeneratorBasedBuilder):
                 },
             ),
             datasets.SplitGenerator(
-                name=datasets.Split.VALIDATION,
+                name=datasets.Split.TEST,
                 # These kwargs will be passed to _generate_examples
                 gen_kwargs={
-                    "claims": claims_validation,
+                    "claims": claims_test,
                     "corpus": corpus,
-                    "split": "validation",
+                    "split": "test",
                 },
             ),
         ]
@@ -127,7 +131,7 @@ class ScifactEntailment(datasets.GeneratorBasedBuilder):
         id_ = -1  # Will increment to 0 on first iteration.
         for claim in claims:
             evidence = {int(k): v for k, v in claim["evidence"].items()}
-            for cited_doc_id in claim["cited_doc_ids"]:
+            for cited_doc_id in claim["doc_ids"]:
                 cited_doc = corpus[cited_doc_id]
                 abstract_sents = [sent.strip() for sent in cited_doc["abstract"]]