ACORD Clause Retrieval Dataset

A ACORD Clause Retrieval Dataset is a graded-relevance expert-annotated legal clause dataset that contains relevance annotations for contract clause retrieval tasks to support precedent-based contract drafting.

• AKA: Atticus Clause Retrieval Dataset, ACORD Dataset, Clause Retrieval Corpus, Precedent Clause Dataset.
• Context:
  • It can typically contain 114 Querys paired with 126,000+ Query-Clause Pairs from legal contracts.
  • It can typically include 400+ EDGAR Contracts as source documents for clause extraction.
  • It can typically provide 1-5 Star Ratings indicating clause quality and relevance levels.
  • It can typically support Graded Relevance Evaluation using NDCG metrics.
  • It can typically enable Retrieval-Augmented Generation for contract drafting applications.
  • ...
  • It can often facilitate Few-Shot Clause Learning through high-quality examples.
  • It can often benchmark Dense Retrieval Systems against sparse baselines.
  • It can often evaluate Cross-Domain Generalization across contract types.
  • It can often measure Precision at K for top-ranked clauses.
  • ...
  • It can range from being a Small ACORD Clause Retrieval Dataset to being a Large ACORD Clause Retrieval Dataset, depending on its dataset size.
  • It can range from being a Single-Domain ACORD Clause Retrieval Dataset to being a Multi-Domain ACORD Clause Retrieval Dataset, depending on its contract domain coverage.
  • It can range from being a Binary ACORD Clause Retrieval Dataset to being a Graded ACORD Clause Retrieval Dataset, depending on its relevance annotation granularity.
  • It can range from being a Static ACORD Clause Retrieval Dataset to being a Evolving ACORD Clause Retrieval Dataset, depending on its update frequency.
  • ...
  • It can integrate with Clause Similarity Retrieval Mechanisms for system training.
  • It can support Retrieval-Augmented Natural Language Generation (RAG) Techniques for clause generation.
  • It can benchmark Contract Clause Discovery Tasks through retrieval evaluation.
  • It can connect to Contract Management Platforms for clause library building.
  • It can interface with AI-based Contract Review Systems for precedent access.
  • ...
• Example(s):
  • ACORD v1 (2024), with 114 querys and 126,000 pairs from EDGAR filings.
  • Query Types in ACORD, such as:
    • Obligation Querys seeking duty clauses.
    • Right Querys finding entitlement provisions.
    • Restriction Querys locating limitation clauses.
  • Annotation Levels in ACORD, such as:
    • 5-Star Clauses representing perfect precedents.
    • 3-Star Clauses showing partial relevance.
    • 1-Star Clauses indicating minimal applicability.
  • ...
• Counter-Example(s):
  • Binary Classification Datasets like CUAD, which lack graded relevance.
  • Document-Level Datasets without clause-level annotations.
  • Synthetic Legal Datasets lacking real contract examples.
• See: Contract Understanding Atticus Dataset (CUAD), LEDGAR Dataset, Clause Similarity Retrieval Mechanism, Contract Clause Discovery Task, Retrieval-Augmented Natural Language Generation (RAG) Technique, Legal Clause Identification Task, Annotated Legal Dataset, Contract Drafting System.

---