romanize-string/dist/esm/transliterators/korean-romanization.js

A fully typed, general-purpose utility for unidirectional string transliteration (non-Latin script => Latin script).

import { normalize as koreanNormalize, tokenize as koreanTokenize, } from "oktjs";
import { romanize } from "../vendor/romanize/korean/src/index.js";
export const romanizeKorean = (string) => {
    // Normalize and tokenize string, omitting any extra white spaces
    const normalized = koreanNormalize(string);
    const tokens = koreanTokenize(normalized).filter((t) => t.text.trim());
    // Romanize each token individually
    const romanizedTokens = tokens.map((token) => romanize(token.text));
    // Use the index of each romanizedToken to access the pos property of the corresponding item in the tokens array.
    // If the token at the current index is a josa (Korean case marker), join it to the preceding noun with a hyphen.
    const josaJoined = romanizedTokens.reduce((acc, romToken, idx) => {
        if (tokens[idx].pos === "Josa" && acc.length) {
            acc[acc.length - 1] += `-${romToken}`;
        }
        else {
            acc.push(romToken);
        }
        return acc;
    }, []);
    // Join the processed array of tokens into a single string, omitting any additional extra white spaces
    // that may have crept in during the process and removing white spaces before punctuation.
    return josaJoined
        .join(" ")
        .replace(/\s+/g, " ")
        .replace(/\s+([.,!?！？。、])/g, "$1");
};