---
dataset_info:
  features:
  - name: sentence1
    dtype: string
  - name: sentence2
    dtype: string
  - name: labels
    dtype: string
  - name: metadata
    dtype: string
  splits:
  - name: train
    num_bytes: 83206152
    num_examples: 58529
  - name: validation
    num_bytes: 8918769
    num_examples: 6142
  - name: test
    num_bytes: 88085535
    num_examples: 61988
  download_size: 59636625
  dataset_size: 180210456
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
license: cc-by-4.0
source_datasets:
- theatticusproject/acord
---

# acord

ACORD attorney-annotated query/clause relevance, preserving native query splits.

Join only explicit qrels against the official queries/corpus. Unjudged pairs
are never converted to negatives. Grades 0–4 correspond to 1–5 attorney stars.
Original dataset terms: CC BY 4.0. Archive, source revision and counts are
recorded in provenance.json; metadata retains query and clause identities.

Original data: [theatticusproject/acord](https://huggingface.co/datasets/theatticusproject/acord). Repackaged as parquet for [tasksource](https://github.com/sileod/tasksource) by [scripts/repackage_dataset/specialized_text.py](https://github.com/sileod/tasksource/blob/main/scripts/repackage_dataset/specialized_text.py#L23).

[Conversion provenance](provenance.json): source revisions, hashes, configuration and exclusions.