Datasetpublic

renzzyyy1028/civil-code-phil

Civilex — Philippine Legal RAG & SFT Dataset Retrieval corpus and supervised fine-tuning (SFT) data for a retrieval-augmented generation (RAG) pipeline over Philippine law: the Civil Code (Republic Act No. 386) and Supreme Court jurisprudence. Produced by the civilex-thesis research pipeline. Contents: 11k+ Supreme Court jurisprudence cases spanning 1949–2025, and 2,270 articles from the Civil Code (Republic Act No. 386). Dataset structure . ├── README.md ├──… See the full description on the dataset page: https://huggingface.co/datasets/renzzyyy1028/civil-code-phil.

sourceHugging Facecc-by-4.0updated 5h agoView on Hugging Face

0likes144downloads

Dataset Card

Civilex — Philippine Legal RAG & SFT Dataset

Retrieval corpus and supervised fine-tuning (SFT) data for a retrieval-augmented generation (RAG) pipeline over Philippine law: the Civil Code (Republic Act No. 386) and Supreme Court jurisprudence. Produced by the civilex-thesis research pipeline.

Contents: 11k+ Supreme Court jurisprudence cases spanning 1949–2025, and 2,270 articles from the Civil Code (Republic Act No. 386).

Dataset structure

.
├── README.md
├── civil_code_rag.jsonl          # Civil Code articles + hierarchy + citation linkage
├── jurisprudence_chunks.jsonl    # RAG-ready chunks of Supreme Court decisions
├── jurisprudence_document.jsonl  # Full Supreme Court decisions (one row per case)
├── article_case_index.json       # Auxiliary: article → case citation index (dict, not a table)
└── training/
    ├── train_clean.jsonl         # SFT chat examples (train)
    └── eval.jsonl                # SFT chat examples (validation)

Subsets (configs)

Config	File	Rows	Description
`civil_code`	`civil_code_rag.jsonl`	2,270	Civil Code articles with hierarchy, cleaned text, topics, keywords, cross-references, and per-article jurisprudence linkage
`jurisprudence_chunks`	`jurisprudence_chunks.jsonl`	351,076	Chunked decision text for retrieval; each chunk links back to its case and cited Civil Code articles
`jurisprudence_documents`	`jurisprudence_document.jsonl`	11,879	Full decisions: metadata, content, summary, dispositive portion, citations
`training`	`training/*.jsonl`	4,351 train / 485 validation	Chat-format SFT examples (system/user/assistant) for plain-language explanation of legal provisions

Data fields

`civil_code`

Field	Type	Description
`article_id`	string	Stable ID, e.g. `RA386-ART1`
`article_number`	int32	Article number
`law_title` / `act_name`	string	e.g. `Republic Act No. 386` / `Civil Code of the Philippines`
`hierarchy`	struct	`{book, book_name, title, title_name, chapter, chapter_name, section, section_name}`
`path`	string	Human-readable location, e.g. `PRELIMINARY TITLE > CH 1 - ... > Art. 1`
`text` / `text_clean`	string	Original and normalized article text
`legal_topics` / `keywords`	list<string>	Topic labels and keywords
`cross_references`	struct	`{cites, cited_by}` article cross-references
`jurisprudence_linkage`	struct	`{total_citations, cases, citation_trend}` per-article case citations
`is_hub`	bool	Whether the article is flagged as a high-citation hub
`chunks`	list<struct>	Pre-computed retrieval chunks for the article
`source_url`	string	Source URL

`jurisprudence_chunks`

Field	Type	Description
`chunk_id`	string	Unique chunk ID (derived from case UID + section)
`case_uid`	string	Join key to `jurisprudence_documents.case_uid`
`gr_number`	string	G.R. number
`year`	int32	Decision year
`legal_topics`	list<string>	Topic labels
`civil_code_articles`	list<int32>	Cited Civil Code article numbers (join key to `civil_code.article_number`)
`section`	string	Source section of the decision the chunk came from
`text`	string	Chunk text (includes a prepended case header line)

`jurisprudence_documents`

Field	Type	Description
`case_id`	string	Case title
`case_uid`	string	Stable internal ID (join key)
`gr_number` / `year` / `decision_date`	string / int32 / string	Case identifiers
`court` / `division` / `case_type`	string	e.g. `Supreme Court`, `EN BANC`, `Appeal`
`url`	string	Source URL (lawphil.net)
`content`	string	Full decision text
`content_summary` / `dispositive_portion`	string	Generated summary and operative portion
`civil_code_citations`	list<struct>	Structured citations into the Civil Code
`legal_topics` / `keywords`	list<string>	Topic labels and keywords
`chunks`	list<struct>	Embedded chunk records

`training`

Field	Type	Description
`id`	string	Example ID, e.g. `T1-EN-RA386-ART133`
`track`	string	Task track (e.g. `simplification`)
`language`	string	`en` / `fil`
`messages`	list<struct>	Chat turns: `{role, content, ...}`
`metadata`	struct	Provenance/track metadata

`article_case_index.json` (auxiliary, not a config)

Top-level dict (not a table): {article_id: {total_jurisprudence_citations, top_cases, ...}}. Not loadable via load_dataset configs — load it directly:

python

import json
index = json.load(open("article_case_index.json"))

Usage

python

from datasets import load_dataset

civil_code = load_dataset("<org>/<repo>", "civil_code")["train"]
chunks     = load_dataset("<org>/<repo>", "jurisprudence_chunks")["train"]
cases      = load_dataset("<org>/<repo>", "jurisprudence_documents")["train"]
sft_train  = load_dataset("<org>/<repo>", "training")["train"]
sft_eval   = load_dataset("<org>/<repo>", "training")["validation"]

Join path for RAG: jurisprudence_chunks.case_uid → jurisprudence_documents.case_uid, and jurisprudence_chunks.civil_code_articles → civil_code.article_number.

Dataset creation

—Statutes: Civil Code of the Philippines (Republic Act No. 386), structured into articles with hierarchy.
—Jurisprudence: Supreme Court decisions sourced from lawphil.net, chunked and annotated with topic labels and Civil Code citations.
—Training data: chat-format examples generated from the corpus for plain-language legal explanation (research use, civilex-thesis pipeline).

Considerations for using the data

—Not legal advice. This is a research corpus for a student thesis; annotations, summaries, and SFT outputs are machine-generated and may contain errors.
—Coverage is limited to the Civil Code and the scraped jurisprudence subset — not a complete corpus of Philippine law.
—Bias: topic labels and summaries are produced by automated pipelines; verify against the source url before relying on any record.
—Privacy: decisions are public court records; case names appear as published.

License

Dataset documentation and annotations: CC BY 4.0. Underlying statutes are laws of the Philippines; court decisions are public records sourced from lawphil.net. Attribution: Civilex thesis research project.