zhconv-rs — 中文简繁及地區詞轉換

zhconv-rs converts Chinese between Traditional, Simplified and regional variants, using rulesets sourced from MediaWiki/Wikipedia and OpenCC, which are merged, flattened and prebuilt into Aho‑Corasick automata for single-pass, linear-time conversions.

🔗 Web app (wasm): https://zhconv.pages.dev (w/ OpenCC dictionaries)

⚙️ Cli: cargo install zhconv or download from releases

🦀 Rust crate: cargo add zhconv (see docs for details)

use zhconv::{zhconv, Variant};
assert_eq!(zhconv("雾失楼台，月迷津渡", Variant::ZhTW), "霧失樓臺，月迷津渡");
assert_eq!(zhconv("驛寄梅花，魚傳尺素", "zh-Hans".parse().unwrap()), "驿寄梅花，鱼传尺素");

🐍 Python package w/ wheels: pip install zhconv-rs or pip install zhconv-rs-opencc (for OpenCC dictionaries)

# > pip install zhconv_rs
# Convert using the built-in rulesets:
from zhconv_rs import zhconv
assert zhconv("天干物燥 小心火烛", "zh-tw") == "天乾物燥 小心火燭"
assert zhconv("《-{zh-hans:三个火枪手;zh-hant:三劍客;zh-tw:三劍客}-》是亞歷山大·仲馬的作品。", "zh-cn", mediawiki=True) == "《三个火枪手》是亚历山大·仲马的作品。"
assert zhconv("-{H|zh-cn:雾都孤儿;zh-tw:孤雛淚;zh-hk:苦海孤雛;zh-sg:雾都孤儿;zh-mo:苦海孤雛;}-《雾都孤儿》是查尔斯·狄更斯的作品。", "zh-tw", True) == "《孤雛淚》是查爾斯·狄更斯的作品。"

# Convert using custom rules:
from zhconv_rs import make_converter
assert make_converter(None, [("天", "地"), ("水", "火")])("甘肅天水") == "甘肅地火"

import io
convert = make_converter("zh-hans", io.StringIO("䖏 处\n罨畫 掩画")) # or path to rule file
assert convert("秀州西去湖州近 幾䖏樓臺罨畫間") == "秀州西去湖州近 几处楼台掩画间"

🧩 API demo: https://zhconv.bamboo.workers.dev

Node.js package: npm install zhconv or yarn add zhconv

JS in browser: https://cdn.jsdelivr.net/npm/zhconv-web@latest

<script type="module">
    // Use ES module import syntax to import functionality from the module
    // that we have compiled.
    //
    // Note that the `default` import is an initialization function which
    // will "boot" the module and make it ready to use. Currently browsers
    // don't support natively imported WebAssembly as an ES module, but
    // eventually the manual initialization won't be required!
    import init, { zhconv } from 'https://cdn.jsdelivr.net/npm/zhconv-web@latest/zhconv.js'; // specify a version tag if in prod

    async function run() {
        await init();

        alert(zhconv(prompt("Text to convert to zh-hans:"), "zh-hans"));
    }

    run();
</script>

Variants and dictionaries

Unlike OpenCC, whose dictionaries are bidirectional (e.g., s2t, tw2s), zhconv-rs follows MediaWiki’s approach and provides one dictionary per target variant:

| Target | Tag | Script | Description | | -------------------------------------- | --------- | ------- | --------------------------------------------- | | Simplified Chinese / 简体中文 | zh-Hans | SC / 简 | W/O substituing region-specific phrases. | | Traditional Chinese / 繁體中文 | zh-Hant | TC / 繁 | W/O substituing region-specific phrases. | | Chinese (Taiwan) / 臺灣正體 | zh-TW | TC / 繁 | With Taiwan-specific phrases adapted. | | Chinese (Hong Kong) / 香港繁體 | zh-HK | TC / 繁 | With Hong Kong-specific phrases adapted. | | Chinese (Macau) / 澳门繁體 | zh-MO | TC / 繁 | Same as zh-HK for now. | | Chinese (Mainland China) / 大陆简体 | zh-CN | SC / 简 | With mainland China-specific phrases adapted. | | Chinese (Singapore) / 新加坡简体 | zh-SG | SC / 简 | Same as zh-CN for now. | | Chinese (Malaysia) / 大马简体 | zh-MY | SC / 简 | Same as zh-CN for now. |

Note: zh-TW and zh-HK are derived from zh-Hant. zh-CN is derived from zh-Hans. Currently, zh-MO shares the same dictionary as zh-HK, and zh-MY/zh-SG share the same dictionary as zh-CN, unless additional rules are provided.

Chained dictionary groups from OpenCC are flattened and merged with MediaWiki dictionaries for each target variant, then compiled into a single Aho-Corasick automaton at build time. After internal compression, the bundled dictionaries and automata occupy ~0.6 MiB (without OpenCC) or ~2.7 MiB (with OpenCC enabled).

Performance

Even with all dictionaries enabled, zhconv-rs remains faster than most alternatives. Check with cargo bench compare --features opencc:

Conversion runs in a single pass in O(n+m) linear time by default, where n is the length of the input text and m is the maximum length of source word in dictionaries, regardless of enabled dictionaries. When converting wikitext containing MediaWiki conversion rules, the time complexity may degrade to O(n*m) in the worst case, if the corresponding function or flag is explicitly chosen.

On a typical modern PC, prebuilt converters load in a few milliseconds with default features (~2–5 ms). Enabling the optional opencc feature increases load time (typically 20–25 ms per target). Throughput generally ranges from 100–200 MB/s.

cargo bench --features opencc on AMD EPYC 7B13 (GitPod) by v0.3:

load/zh2Hant            time:   [4.6368 ms 4.6862 ms 4.7595 ms]
load/zh2Hans            time:   [2.2670 ms 2.2891 ms 2.3138 ms]
load/zh2TW              time:   [4.7115 ms 4.7543 ms 4.8001 ms]
load/zh2HK              time:   [5.4438 ms 5.5474 ms 5.6573 ms]
load/zh2MO              time:   [4.9503 ms 4.9673 ms 4.9850 ms]
load/zh2CN              time:   [3.0809 ms 3.1046 ms 3.1323 ms]
load/zh2SG              time:   [3.0543 ms 3.0637 ms 3.0737 ms]
load/zh2MY              time:   [3.0514 ms 3.0640 ms 3.0787 ms]
zh2CN wikitext basic    time:   [385.95 µs 388.53 µs 391.39 µs]
zh2TW wikitext basic    time:   [393.70 µs 395.16 µs 396.89 µs]
zh2TW wikitext extended time:   [1.5105 ms 1.5186 ms 1.5271 ms]
zh2CN 天乾物燥          time:   [46.970 ns 47.312 ns 47.721 ns]
zh2TW data54k           time:   [200.72 µs 201.54 µs 202.41 µs]
zh2CN data54k           time:   [231.55 µs 232.86 µs 234.30 µs]
zh2Hant data689k        time:   [2.0330 ms 2.0513 ms 2.0745 ms]
zh2TW data689k          time:   [1.9710 ms 1.9790 ms 1.9881 ms]
zh2Hant data3185k       time:   [15.199 ms 15.260 ms 15.332 ms]
zh2TW data3185k         time:   [15.346 ms 15.464 ms 15.629 ms]
zh2TW data55m           time:   [329.54 ms 330.53 ms 331.58 ms]
is_hans data55k         time:   [404.73 µs 407.11 µs 409.59 µs]
infer_variant data55k   time:   [1.0468 ms 1.0515 ms 1.0570 ms]
is_hans data3185k       time:   [22.442 ms 22.589 ms 22.757 ms]
infer_variant data3185k time:   [60.205 ms 60.412 ms 60.627 ms]

load/zh2Hant            time:   [22.074 ms 22.338 ms 22.624 ms]
load/zh2Hans            time:   [2.7913 ms 2.8126 ms 2.8355 ms]
load/zh2TW              time:   [23.068 ms 23.286 ms 23.520 ms]
load/zh2HK              time:   [23.358 ms 23.630 ms 23.929 ms]
load/zh2MO              time:   [23.363 ms 23.627 ms 23.913 ms]
load/zh2CN              time:   [3.6778 ms 3.7222 ms 3.7722 ms]
load/zh2SG              time:   [3.6522 ms 3.6848 ms 3.7202 ms]
load/zh2MY              time:   [3.6642 ms 3.7079 ms 3.7545 ms]
zh2CN wikitext basic    time:   [396.17 µs 402.51 µs 409.36 µs]
zh2TW wikitext basic    time:   [442.16 µs 447.53 µs 453.27 µs]
zh2TW wikitext extended time:   [1.5795 ms 1.6007 ms 1.6233 ms]
zh2CN 天乾物燥          time:   [47.884 ns 48.878 ns 49.953 ns]
zh2TW data54k           time:   [255.25 µs 259.01 µs 262.92 µs]
zh2CN data54k           time:   [233.74 µs 236.99 µs 240.67 µs]
zh2Hant data689k        time:   [3.9696 ms 4.0005 ms 4.0327 ms]
zh2TW data689k          time:   [3.4593 ms 3.4896 ms 3.5203 ms]
zh2Hant data3185k       time:   [27.710 ms 27.955 ms 28.206 ms]
zh2TW data3185k         time:   [30.298 ms 30.858 ms 31.428 ms]
zh2TW data55m           time:   [500.95 ms 515.80 ms 531.34 ms]
is_hans data55k         time:   [461.22 µs 470.99 µs 481.20 µs]
infer_variant data55k   time:   [1.1669 ms 1.1759 ms 1.1852 ms]
is_hans data3185k       time:   [26.609 ms 26.964 ms 27.385 ms]
infer_variant data3185k time:   [74.878 ms 76.262 ms 77.818 ms]

Limitations

Accuracy

Rule-based converters cannot capture every possible linguistic nuance. Like most others, the implementation employs a leftmost-longest matching strategy (a.k.a forward maximum matching), prioritizing to the earliest and longest matches in the text. For example, if a ruleset contains both 干 → 幹, 天干 → 天干, and 天干物燥 → 天乾物燥, the converter will prefer the longer match 天乾物燥, since it appears earlier and spans more characters. This generally works well but may cause occasional mis-conversions.

Wikitext support

The implementation supports most MediaWiki conversion rules, while not fully compliant with the original MediaWiki implementation.

Since rebuilding automata dynamically is impractical, rules (e.g., -{H|zh-hans:鹿|zh-hant:马}- in MediaWiki syntax) in text are extracted in a first pass, a temporary automaton is constructed, and the text is converted in a second pass. The time complexity may degrade to O(n*m) in the worst case, where n is the input text length and m is the maximum length of source words in dictionaries, which is equivalent to a brute-force approach.

Credits

Rulesets/Dictionaries: MediaWiki and OpenCC.

Fast double-array Aho-Corasick automata implementation in Rust: daachorse

References & related implementations:

• https://github.com/gumblex/zhconv : Python implementation of zhConver{ter,sion}.php.
• https://github.com/BYVoid/OpenCC/ : Widely adopted Chinese converter.
• https://zh.wikipedia.org/wiki/Wikipedia:字詞轉換處理
• https://zh.wikipedia.org/wiki/Help:高级字词转换语法
• https://github.com/wikimedia/mediawiki/blob/master/includes/language/LanguageConverter.php