Dataset Structure
This dataset contains two subsets:
QA: Question-answer pairs, spanning both single-turn and multi-turn interactions
conversation_id(string): A unique identifier for a conversation session. In multi-turn configurations, multiple rows share the same ID to represent a continuous dialogue.turn_id(int32): The sequential order of messages within a session (0represents the first user query).question(string): The question text.ground_truth(string): The reference response.
Documents: Document contents referenced by the QA subset
id(string): Unique document identifier.content(string): The document text content.
Data Construction
The data is constructed using Expert-Crafted Data; Questions and their corresponding reference answers are crafted by domain experts in the legal and judicial field. Each interaction is manually written to reflect realistic single-turn and multi-turn conversational scenarios. The reference documents are drawn from legal reference texts and statutory laws.
Source
| Subset | Source |
|---|---|
| QA | Expert-crafted single-turn and multi-turn conversations based on legal statutes and regulations |
| Documents | Legal statutes, regulations, and reference texts |
Review Process
All data undergoes a manual human review process. Problematic samples are directly removed or modified while preserving their original intent. Reviewers may also use automated tools to assist in this process.
| # | Criterion | Description |
|---|---|---|
| 1 | Factual Accuracy | The ground truth response must be legally accurate. |
| 2 | Conversational Coherence | In multi-turn settings, each turn must flow naturally from the preceding context without contradiction or redundancy. |
| 3 | Completeness and Clarity | Each question and answer must be self-contained within its conversation context and free of ambiguity. |
- Downloads last month
- 5