kamus

A command-line tool for looking up, searching, and exporting the Kamus Dewan Malay dictionary dataset (converted from PDF to JSON, ~52,212 entries).

Data source & schema

The dataset is a flat JSON array of word entries, originally extracted from a PDF of Kamus Dewan. Each entry looks roughly like:

{
  "word": "abadi",
  "origin": "Ar",
  "rootWord": "abad",
  "aliases": ["abadi-abadi"],
  "definitions": [
    { "sense": 1, "text": "..." },
    { "sense": 2, "text": "..." }
  ]
}

Schema quirks this tool accounts for:

• Numbered homonym variants — words like "a I" / "a II" or "aba I" / "aba II" disambiguate unrelated words that happen to be spelled the same. kamus lookup shows all variants together when you search the bare word (a, aba), and shows only the requested one if you include the suffix explicitly ("aba II").
• rootWord — links a derived form back to its base word (e.g. mengabadikan → abadi). kamus root uses this to reconstruct full derivation families.
• aliases — related/alternate forms, shown alongside a lookup result but not used for indexing.
• sense — some definitions are numbered (multiple senses), some aren't (single sense, just a text field). Both are rendered appropriately.
• origin — present for loanwords (Ar = Arabic, Id = Indonesian, etc.), absent otherwise.

Place the full 52K-entry dataset at data/kamus.json (a small sample is included there for testing — replace it with the full file).

Installation

git clone <this repo>
cd kamus-cli
pip install -e .

This installs the kamus command globally (within your active Python environment) via the project.scripts entry point in pyproject.toml.

Usage

kamus lookup <word>

Exact lookup. Shows origin, root word, aliases, and all definitions. Numbered homonym variants are shown together if you omit the suffix. Falls back to fuzzy (rapidfuzz) suggestions if there's no exact match.

kamus lookup abadi
kamus lookup aba          # shows both "aba I" and "aba II"
kamus lookup "aba II"     # shows only that one variant
kamus lookup abadii       # no match -> "did you mean: abadi, abad, ..."

kamus random

Word of the day — a random entry with its full definition.

kamus random

kamus search <term>

Substring match across all words. Returns a word list only (definitions are omitted, since this can match many entries).

kamus search abad

kamus root <word>

Shows a word's root word (if any) and every other word in the dataset sharing that root — the full derivation family.

kamus root mengabadikan
# Root word: abadi
# Derivation family (3 other form(s)):
#   - terabadi
#   - keabadian
#   - pengabadian

kamus root abadi          # works from the root itself too

kamus export --format <csv|json|sqlite> --output <path>

Exports the full dataset.

• json: the raw nested dataset, unflattened, written as-is.
• csv / sqlite: one row per definition (sense), not per word, since definitions is a nested array. Columns: word, origin, root_word, aliases, sense, definition_text. A word with 2 senses produces 2 rows sharing the same word/origin/etc. See src/kamus/export.py for the full rationale.

kamus export --format csv --output kamus.csv
kamus export --format sqlite --output kamus.sqlite
kamus export --format json --output kamus_export.json

--json on every command

Every command accepts --json to print raw JSON instead of formatted output, for scripting/piping:

kamus lookup abadi --json | jq '.[0].definitions'
kamus search abad --json

How the data is loaded

The full JSON file is loaded into memory once per process and indexed (by exact word, by base word for homonym grouping, by root word for derivation families) at startup. At ~52K entries this takes a fraction of a second and keeps the implementation simple — see the docstring at the top of src/kamus/data.py for the full reasoning, including when a heavier on-disk index would actually be worth it.

Tests

pip install -e ".[test]"
pytest

Tests run against a small fixture (tests/fixtures/sample.json) drawn from the real dataset's first ~14 entries, covering exact lookup, homonym-variant handling, fuzzy-match fallback, root-word derivation families, and export correctness (csv/json/sqlite).

Project structure

kamus-cli/
├── pyproject.toml
├── README.md
├── data/
│   └── kamus.json        # full dataset goes here
├── src/kamus/
│   ├── __init__.py
│   ├── data.py            # loading + indexing
│   ├── export.py          # csv/json/sqlite export
│   └── cli.py             # click commands
└── tests/
    ├── fixtures/sample.json
    ├── test_lookup.py
    ├── test_root.py
    └── test_export.py