mirror of
https://github.com/ksyasuda/SubMiner.git
synced 2026-02-27 18:22:41 -08:00
179 lines
4.5 KiB
TypeScript
179 lines
4.5 KiB
TypeScript
/*
|
|
* SubMiner - All-in-one sentence mining overlay
|
|
* Copyright (C) 2024 sudacode
|
|
*
|
|
* This program is free software: you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation, either version 3 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program. If not, see <https://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
import { spawn, execSync } from "child_process";
|
|
import { PartOfSpeech, Token, MecabStatus } from "./types";
|
|
import { createLogger } from "./logger";
|
|
|
|
export { PartOfSpeech };
|
|
|
|
const log = createLogger("mecab");
|
|
|
|
function mapPartOfSpeech(pos1: string): PartOfSpeech {
|
|
switch (pos1) {
|
|
case "名詞":
|
|
return PartOfSpeech.noun;
|
|
case "動詞":
|
|
return PartOfSpeech.verb;
|
|
case "形容詞":
|
|
return PartOfSpeech.i_adjective;
|
|
case "形状詞":
|
|
case "形容動詞":
|
|
return PartOfSpeech.na_adjective;
|
|
case "助詞":
|
|
return PartOfSpeech.particle;
|
|
case "助動詞":
|
|
return PartOfSpeech.bound_auxiliary;
|
|
case "記号":
|
|
case "補助記号":
|
|
return PartOfSpeech.symbol;
|
|
default:
|
|
return PartOfSpeech.other;
|
|
}
|
|
}
|
|
|
|
export function parseMecabLine(line: string): Token | null {
|
|
if (!line || line === "EOS" || line.trim() === "") {
|
|
return null;
|
|
}
|
|
|
|
const tabIndex = line.indexOf("\t");
|
|
if (tabIndex === -1) {
|
|
return null;
|
|
}
|
|
|
|
const surface = line.substring(0, tabIndex);
|
|
const featureString = line.substring(tabIndex + 1);
|
|
const features = featureString.split(",");
|
|
|
|
const pos1 = features[0] || "";
|
|
const pos2 = features[1] || "";
|
|
const pos3 = features[2] || "";
|
|
const pos4 = features[3] || "";
|
|
const inflectionType = features[4] || "";
|
|
const inflectionForm = features[5] || "";
|
|
const lemma = features[6] || surface;
|
|
const reading = features[7] || "";
|
|
const pronunciation = features[8] || "";
|
|
|
|
return {
|
|
word: surface,
|
|
partOfSpeech: mapPartOfSpeech(pos1),
|
|
pos1,
|
|
pos2,
|
|
pos3,
|
|
pos4,
|
|
inflectionType,
|
|
inflectionForm,
|
|
headword: lemma !== "*" ? lemma : surface,
|
|
katakanaReading: reading !== "*" ? reading : "",
|
|
pronunciation: pronunciation !== "*" ? pronunciation : "",
|
|
};
|
|
}
|
|
|
|
export class MecabTokenizer {
|
|
private mecabPath: string | null = null;
|
|
private available: boolean = false;
|
|
private enabled: boolean = true;
|
|
|
|
async checkAvailability(): Promise<boolean> {
|
|
try {
|
|
const result = execSync("which mecab", { encoding: "utf-8" }).trim();
|
|
if (result) {
|
|
this.mecabPath = result;
|
|
this.available = true;
|
|
log.info("MeCab found at:", this.mecabPath);
|
|
return true;
|
|
}
|
|
} catch (err) {
|
|
log.info("MeCab not found on system");
|
|
}
|
|
|
|
this.available = false;
|
|
return false;
|
|
}
|
|
|
|
async tokenize(text: string): Promise<Token[] | null> {
|
|
if (!this.available || !this.enabled || !text) {
|
|
return null;
|
|
}
|
|
|
|
return new Promise((resolve) => {
|
|
const mecab = spawn("mecab", [], {
|
|
stdio: ["pipe", "pipe", "pipe"],
|
|
});
|
|
|
|
let stdout = "";
|
|
let stderr = "";
|
|
|
|
mecab.stdout.on("data", (data: Buffer) => {
|
|
stdout += data.toString();
|
|
});
|
|
|
|
mecab.stderr.on("data", (data: Buffer) => {
|
|
stderr += data.toString();
|
|
});
|
|
|
|
mecab.on("close", (code: number | null) => {
|
|
if (code !== 0) {
|
|
log.error("MeCab process exited with code:", code);
|
|
if (stderr) {
|
|
log.error("MeCab stderr:", stderr);
|
|
}
|
|
resolve(null);
|
|
return;
|
|
}
|
|
|
|
const lines = stdout.split("\n");
|
|
const tokens: Token[] = [];
|
|
|
|
for (const line of lines) {
|
|
const token = parseMecabLine(line);
|
|
if (token) {
|
|
tokens.push(token);
|
|
}
|
|
}
|
|
|
|
resolve(tokens);
|
|
});
|
|
|
|
mecab.on("error", (err: Error) => {
|
|
log.error("Failed to spawn MeCab:", err.message);
|
|
resolve(null);
|
|
});
|
|
|
|
mecab.stdin.write(text);
|
|
mecab.stdin.end();
|
|
});
|
|
}
|
|
|
|
getStatus(): MecabStatus {
|
|
return {
|
|
available: this.available,
|
|
enabled: this.enabled,
|
|
path: this.mecabPath,
|
|
};
|
|
}
|
|
|
|
setEnabled(enabled: boolean): void {
|
|
this.enabled = enabled;
|
|
}
|
|
}
|
|
|
|
export { mapPartOfSpeech };
|