JavaScript implementation of Japanese morphological analyzer. This is a pure JavaScript porting of Kuromoji.
You can see how kuromoji.js works in demo site.
Directory tree is as follows:
build/
kuromoji.js -- JavaScript file for browser (Browserified)
demo/ -- Demo
dict/ -- Dictionaries for tokenizer (uncompressed)
example/ -- Examples to use in Node.js
src/ -- JavaScript source
test/ -- Unit test
You can tokenize sentences with only 5 lines of code. If you need working examples, you can see the files under the demo or example directory.
Install with npm package manager:
npm install kuromoji
Load this library as follows:
var kuromoji = require("kuromoji");
You can prepare tokenizer like this:
kuromoji.loader({ dic_path: "path/to/dictionary/dir/" }).load(function (err, tokenizer) {
// tokenizer is ready
var path = tokenizer.tokenize("すもももももももものうち");
console.log(path);
});
You only need the build/kuromoji.js and dict/*.dat files
In your JavaScript:
kuromoji.loader({ dic_path: "/url/to/dictionary/dir/" }).load(function (err, tokenizer) {
// tokenizer is ready
var path = tokenizer.tokenize("すもももももももものうち");
console.log(path);
});
On your server:
Host the dictionary files and compress them using either gzip or brotli. Make sure to
set the `Content-Type` header so they are automatically decompressed by the browser.
The function tokenize() returns an JSON array like this:
[ {
word_id: 509800, // 辞書内での単語ID
word_type: 'KNOWN', // 単語タイプ(辞書に登録されている単語ならKNOWN, 未知語ならUNKNOWN)
word_position: 1, // 単語の開始位置
surface_form: '黒文字', // 表層形
pos: '名詞', // 品詞
pos_detail_1: '一般', // 品詞細分類1
pos_detail_2: '*', // 品詞細分類2
pos_detail_3: '*', // 品詞細分類3
conjugated_type: '*', // 活用型
conjugated_form: '*', // 活用形
basic_form: '黒文字', // 基本形
reading: 'クロモジ', // 読み
pronunciation: 'クロモジ' // 発音
} ]
(This is defined in src/util/IpadicFormatter.js)
See also JSDoc page in details.