Datasets:
The dataset viewer is not available for this split.
Error code: TooBigContentError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Philippine jurisprudence & statutes (law_ph) - government works
Supreme Court decisions (juris) and statutes (statutes - Republic and
Commonwealth Acts, Batas Pambansa, Presidential Decrees) of the Philippines,
1900-2025, re-derived from bettergovph/gov-library (HTML->Markdown extraction
of LawPhil.net, Arellano Law Foundation) with per-record
provenance (LawPhil path/URL, G.R. number, date).
- Source rows in scope: 87,873 (upstream total 114,340 incl. executive issuances and constitutions - out of scope here)
- Retained records: 85,534 (97.3%) -
juris65,843 +statutes19,695 - Measured
cl100k_baseproxy tokens: 346,701,560 - Characters: 1,472,572,736
- Time span (document dates): 1900 - 2025-07-01 (year precision; month where available)
- Language reality: the Philippine legal register is English-dominant -
candidate language votes: 85,586
en, 31tl. Per-record detected language is inartifacts/attribution.jsonl; this is a Philippine-legal corpus, not Filipino-language practice text.
License - evidence at the origin
- Government works: RA 8293 (Intellectual Property Code of the
Philippines), s. 176.1 - "No copyright shall subsist in any work of the
Government of the Philippines." Statutes and court decisions are such works.
Proviso to note: commercial exploitation of government works requires prior
approval of the originating agency (s. 176) - a usage condition, not a
copyright. Per-record license field:
LicenseRef-PD-Official. - Chain-of-title gaps, stated plainly: LawPhil.net (Arellano Law
Foundation) is a digitization mirror of government texts, not the rights
source; the
bettergovph/gov-librarypackaging declares no licence at all (its README## Licensesection is empty) and its scraping/extraction code is not published with the package.release: Nonepending the standard rights review.
Samples
Verbatim records from artifacts/sample.jsonl, real line breaks preserved
(markdown footnote markers from the HTML extraction are kept as delivered),
truncated with [...].
Sample 1
id: law_ph_juris-datasets-markdown-juris-2000-10-gr-117949-2000-
source: law_ph
created: 2000-10-01
token_count: 5909
author: Supreme Court of the Philippines
**G.R. No. 117949 October 23, 2000**
PEOPLE OF THE PHILIPPINES, plaintiff-appellee, vs. **ALEX BANTILLO, ERNESTO ASUNCION, LARRY ASUNCION, RODOLFO ATANAS, PAQUITO FERNANDEZ and ANORING BADANDO,** accused, **ALEX BANTILLO, and ERNESTO ASUNCION,** accused-appellants.
**QUISUMBING, *J.:***
This is an appeal from the decision[1](#fn [...]
Sample 2
id: law_ph_juris-datasets-markdown-juris-1928-03-gr-28050-1928-m
source: law_ph
created: 1928-03-01
token_count: 2567
author: Supreme Court of the Philippines
**G.R. No. 28050 March 13, 1928**
FEDERICO VALERA, plaintiff-appellant, vs. **MIGUEL VELASCO,** defendant-appellee.
*Jose Martinez San Agustin for appellant. Vicente O. Romualdez, Crispulo T. Manubay and Placido P. Reyes for appellee.*
**VILLA-REAL, *J.:***
This is an appeal taken by Federico Valera from the judgment of the Cour [...]
Sample 3
id: law_ph_juris-datasets-markdown-juris-2016-05-gr-183129-2016-
source: law_ph
created: 2016-05-01
token_count: 8559
author: Supreme Court of the Philippines
**May 30, 2016**
G.R. No. 183129
COCOPLANS, INC. and CAESAR T. MICHELENA, Petitioners, vs. **MA. SOCORRO R. VILLAPANDO,** Respondent.
PERALTA, *J.:*
Before the Court is a petition for review on certiorari under Rule 45 of the Rules of Court seeking to reverse and set aside the Decision[1](#fnt1) dated February 4, 2008 [...]
Filtering
From upstream consolidated.parquet rows (categories juris + statutes):
whitespace normalization -> labelled-only PII redaction (e-mail, labelled phone,
labelled IP, labelled PhilSys/PhilID/TIN/SSS/GSIS/passport identifiers, PH
account candidates: 53 e-mails and 49 labelled identifiers replaced; bare
dotted quads and case numbers deliberately untouched) -> language gate
(langid on first 1,000 chars; keep en/tl) -> cross-document boilerplate
line filter -> exact + near dedup (sha256 of normalized text; Jaccard >= 0.90
on capped 25 x 5-word shingle sets) -> duplicate-id guard.
Rejections (sum exactly to 2,339; one decision per row in
artifacts/decisions.jsonl): 1,312 out of language scope / 673 low
letter-ratio / 215 duplicate upstream ids / 65 too short (min 200 chars) / 41
exact duplicates / 33 near-duplicates.
Known limitations
- English-dominant register (see above);
tlrecords are rare. - Markdown extraction residue: footnote markers
[1](#fn...)and emphasis markers survive in a share of records. - Judgments name private persons (parties, attorneys, witnesses) - not anonymized; a SlayerLab/NERGAL-style review is required before release.
- Executive issuances and constitutions are in the upstream package but out of scope here.
createdis year (month where the upstream records one); 215 rows shared upstream ids with different content and were rejected asduplicate_id.
See artifacts/ for attribution, per-record decisions, QA counts, samples and
checksums. verify() passes (85,534 texts reconstructed from the extracted
source, token counts recomputed); tests 6/6.
- Downloads last month
- 36