README.md

metadata

dataset_info:
  - config_name: distractors
    features:
      - name: evidence
        dtype: string
      - name: evidence_id
        dtype: int64
    splits:
      - name: train
        num_bytes: 735316804
        num_examples: 500000
    download_size: 417661627
    dataset_size: 735316804
  - config_name: test
    features:
      - name: claim
        dtype: string
      - name: evidence
        dtype: string
      - name: evidence_id
        dtype: int64
      - name: label
        dtype: string
      - name: evidences
        sequence: string
      - name: evidence_ids
        sequence: string
      - name: labels
        sequence: string
    splits:
      - name: train
        num_bytes: 1174731
        num_examples: 206
    download_size: 556372
    dataset_size: 1174731
configs:
  - config_name: distractors
    data_files:
      - split: train
        path: distractors/train-*
  - config_name: test
    data_files:
      - split: train
        path: test/train-*
size_categories:
  - 100K<n<1M

Data Stats

• 206 claims
• 500k distractors

Data Structure

Test

• claim
• evidence: GT evidence
• evidence_id: GT evidence id
• label: GT label
• evidences: list of all evidences
• evidence_ids: list of all evidence ids
• labels: list of all labels

Distractors

• evidence
• evidence_id

Process Code

import pandas as pd

from datasets import Dataset

claims = pd.read_csv("./scifact_open_retriever_test.csv")
claims.head()

docs = pd.read_csv("./scifact_open_docs.csv")
docs.head()

id2doc = dict(zip(docs["ID"], docs["Doc"]))

data = {
    "claim": [],
    "evidence": [],
    "evidence_id": [],
    "label": [],
    "evidences": [],
    "evidence_ids": [],
    "labels": [],
}

for i, row in claims.iterrows():
    data["claim"].append(row["Query"])
    evidence_ids = eval(row["Gold"])
    evidence_id = int(evidence_ids[0])
    labels = eval(row["Label"])
    label = str(labels[0])
    data["evidence"].append(id2doc[evidence_id])
    data["evidence_id"].append(evidence_id)
    data["label"].append(label)
    data["evidences"].append([id2doc[int(eid)] for eid in evidence_ids])
    data["evidence_ids"].append(evidence_ids)
    data["labels"].append(labels)

ds = Dataset.from_dict(data)

distractors = {
    "evidence": [],
    "evidence_id": [],
}

for i, row in docs.iterrows():
    distractors["evidence"].append(row["Doc"])
    distractors["evidence_id"].append(row["ID"])

distractors = Dataset.from_dict(distractors)

distractors.push_to_hub("umbc-scify/scifact-open", "distractors")
ds.push_to_hub("umbc-scify/scifact-open", "test")