license: cc-by-4.0
task_categories:
- text-classification
- summarization
- text-generation
language:
- en
tags:
- finance
- insurance
- reinsurance
- contracts
pretty_name: Reinsurance Contract Classification
size_categories:
- 1K<n<10K
configs:
- config_name: default
data_files:
- split: train
path: reinsurance-contracts-2001-2024.parquet
- config_name: extended
data_files:
- split: train
path: reinsurance-contracts-extended-2001-2024.parquet
This dataset contains the full text and the classification of 6671 reinsurance-related documents extracted from SEC filings. The documents include reinsurance contracts, but also ancillary documents, such as amendments, endorsements, extensions, and other documents referring to reinsurance agreement.
The documents have been classified by three language models: qwen3-235b-a22b-2507, gpt-oss-120b and gemini-2.5-flash-lite. Manual inspection of the classification results reveals that qwen3 is generally the most reliable model, followed by gpt-oss. Therefore, the final classification is based on qwen3 classification if at least one of the two other models agree (with some other minor adjustments when some model fail to give an answer).
The classification process identify 990 documents that are obligatory reinsurance contracts, among which about 87% are Non-Life, 12% Life and 1% unknown.
The datasets contains two subsets (or 'configs' in the Datasets package specific terminology). The "default" subset includes only the 990 proper reinsurance contracts. The "extended" subset includes all documents.