A multilingual lexicon dataset of 202,912 entries derived from Sastra.org, a digital archive of Javanese literary heritage maintained by Yayasan Sastra Lestari.

The dataset compiles 32 historical lexicographic sources spanning from the 1830s to 2010, covering Javanese, Old Javanese (Kawi), Indonesian, Dutch, English, and French.

Dataset Structure

The dataset is provided as JSONL files. The full corpus is in data/leksikon_all.jsonl, and is also split by language pair under data/by_pair/ for convenience.

Files

File	Entries	Description
`data/leksikon_all.jsonl`	202,912	Full cleaned corpus
`data/by_pair/ind_jav.jsonl`	11,040	Indonesian ↔ Javanese
`data/by_pair/kawi_ind.jsonl`	12,911	Old Javanese (Kawi) → Indonesian
`data/by_pair/kawi_jav.jsonl`	47,732	Kawi ↔ Javanese
`data/by_pair/jav_jav.jsonl`	49,901	Javanese ↔ Javanese (monolingual)
`data/by_pair/jav_nl.jsonl`	47,926	Javanese ↔ Dutch
`data/by_pair/jav_en.jsonl`	21,531	Javanese ↔ English
`data/by_pair/jav_fr.jsonl`	11,871	Javanese ↔ French

Schema

Field	Type	Description
`id`	int	Original entry ID from source
`headword`	string	Dictionary headword
`lang`	string	Language tag from source: `Ind` (Indonesian→Javanese), `Wngs` (Wangsalan), or empty
`definition`	string	Meaning or translation
`source`	string	Bibliographic source reference with catalog number

Usage

from datasets import load_dataset

# Load the full corpus
ds = load_dataset("junwatu/javanese-multilingual-lexicon", "all")

# Or load a specific language pair
ind_jav = load_dataset("junwatu/javanese-multilingual-lexicon", "ind_jav")
kawi_ind = load_dataset("junwatu/javanese-multilingual-lexicon", "kawi_ind")

Example Entries

{"id": 192902, "headword": "'abdi", "lang": "Ind", "definition": "abdi, batur.", "source": "Bausastra Indonesia-Jawi, Purwadarminta, c. 1939, #1979."}
{"id": 192947, "headword": "'adil", "lang": "Ind", "definition": "adil, di'adili: diadili; peng'adilan: pangadilan.", "source": "Bausastra Indonesia-Jawi, Purwadarminta, c. 1939, #1979."}

Data Processing

The data was cleaned with the following steps:

Deduplicated by (headword, definition) — entries with the same headword but different definitions are preserved
HTML entities decoded (& → &, etc.)
HTML tags stripped
Unicode normalized to NFC form
Whitespace collapsed
Leading : characters removed from definitions

The same headword may appear multiple times across different sources or with different senses. Of 96,063 unique headwords, 46,128 are polysemous (appear with more than one distinct definition).

Sources

The 32 source dictionaries include:

Bausastra Jawa
Bausastra Indonesia-Jawi (Purwadarminta, c. 1939)
Javaansch-Nederlandsch Handwoordenboek (Gericke en Roorda, 1901)
Javaansch-Nederduitsch Woordenboek (Gericke en Roorda, 1847)
Javanese-English Dictionary (Horne, 1974)
Dictionnaire Javanais-Français (L'Abbé P. Favre, 1870)
Têmbung Kawi Mawi Têgêsipun
Kawi - Indonesia (Wojowasito, 1980)
Kawi Lexicon
Wangsalan, Dasanama, and many other smaller works

See the source field in each entry for the full bibliographic reference.

Limitations

Users should be aware of the following characteristics of this dataset before using it:

Historical orthography. Many sources predate modern spelling reforms. Indonesian entries may use pre-1972 EYD spelling (e.g., tjahaja instead of cahaya, oe instead of u). Javanese entries from Dutch-era sources (1847–1901) use colonial romanization conventions that differ from modern Javanese spelling.
Mixed languages within entries. Kawi (Old Javanese) entries often contain Sanskrit etymologies, Latin grammatical notes, and Dutch glosses interleaved with the main definition.
Inconsistent diacritics. Diacritic conventions (ê, è, é, ě, etc.) vary across the 32 source dictionaries. Some sources use phonetic transcription markers in parentheses, e.g., aba (ObO).
Encoded markers. Source-specific abbreviations appear in definitions, such as KN. (krama-ngoko), S. (Sanskrit), Bl. (Balinese), Pm. (proverb). These are not normalized.
OCR-derived noise. Some entries originate from OCR'd scans and may contain occasional character recognition errors.
Not a parallel corpus. Although this dataset includes multiple language pairs, entries are not aligned across sources. The same headword may have different definitions in different dictionaries.
Limited coverage of modern Javanese. Most sources are historical (1830s–1970s). Only a few entries reflect contemporary Javanese usage.

For modern, standardized Indonesian-Javanese translation, consider supplementing this dataset with newer corpora.

Citation

If you use this dataset in your research, please cite both the original source and this dataset:

@misc{sastra_lestari,
  title        = {Sastra.org: Digital Archive of Javanese Literature},
  author       = {{Yayasan Sastra Lestari}},
  howpublished = {\url{https://www.sastra.org/}},
  note         = {Original source content, CC BY-NC 4.0}
}

@misc{javanese_multilingual_lexicon_2026,
  title        = {Javanese Multilingual Lexicon},
  author       = {junwatu},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/junwatu/javanese-multilingual-lexicon}},
  note         = {Derived from Sastra.org under CC BY-NC 4.0}
}

License & Attribution

This dataset contains material from Sastra.org.

Changes made: text extraction, normalization, formatting, metadata generation, dataset structuring.

This dataset is for research and non-commercial use only.

Downloads last month: 9

Total file size:

76.9 MB