---
pretty_name: CODICES Data
tags:
- legal
- multilingual
- parquet
- tabular
configs:
- config_name: with_fulltext
  data_files:
  - split: train
    path: data/with_fulltext/train-00000-of-00001.parquet
- config_name: cases_all
  data_files:
  - split: train
    path: data/cases_all/train-00000-of-00001.parquet
- config_name: precis_only
  data_files:
  - split: train
    path: data/precis_only/train-00000-of-00001.parquet
- config_name: classification
  data_files:
  - split: train
    path: data/classification/train-00000-of-00001.parquet
  - split: validation
    path: data/classification/validation-00000-of-00001.parquet
  - split: test
    path: data/classification/test-00000-of-00001.parquet
---

# CODICES Data

This dataset contains CODICES case metadata, precis fields, full text where available, systematic thesaurus labels, and country-level reference fields.

## Configs

- `with_fulltext`: cases with full text in at least one official language.
- `cases_all`: all cases, including no-fulltext and translation-only records.
- `precis_only`: cases outside the `with_fulltext` bucket.
- `classification`: train, validation, and test splits for classification experiments.

## Legal Origin And Colonial History

`legal_origin`, `colonial_history`, and `klerman_2011_sample` are assigned only from `legal_origin_colonial_history_table2_total.csv`. If a country/entity is not present in that CSV, these fields are left blank.

Allowed `legal_origin` values: `French`, `Common`, `Mixed`, `German`, `Scandinavia`, `Islamic`.

Allowed `colonial_history` values: `French`, `Other French`, `British`, `Austro-Hungarian`, `Other`, `Not Colonized`.

## Reference Files

Reference files are kept under `reference/`. The parquet files remain under `data/` so Hugging Face can render them in the Dataset Viewer.