garu-ko 0.3.1 → 0.3.3

package/dist/index.d.ts

@@ -11,6 +11,8 @@ export interface AnalyzeResult {
     score: number;
     elapsed: number;
 }
+export { normalizeText, splitSentences } from './normalize.js';
+export type { NormalizeOptions, Segment } from './normalize.js';
 export interface AnalyzeOptions {
     topN?: number;
 }

package/dist/index.js

@@ -1,3 +1,4 @@
+export { normalizeText, splitSentences } from './normalize.js';
 const isNode = typeof process !== 'undefined' &&
     process.versions != null &&
     process.versions.node != null;
@@ -132,7 +133,7 @@ export class Garu {
         return {
             version: this._wasm.constructor.version(),
             size: this._modelSize,
-            accuracy: 0.953,
+            accuracy: 0.908,
         };
     }
     /**

package/dist/normalize.d.ts

@@ -0,0 +1,19 @@
+/**
+ * Text normalization utilities for Garu Korean morphological analyzer.
+ * Pre-processes informal/non-standard Korean text before analysis.
+ */
+export interface NormalizeOptions {
+    /** Collapse repeated jamo (ㅋㅋㅋ→ㅋㅋ). Default: true */
+    jamo?: boolean;
+    /** Expand 2-char jamo abbreviations (ㄱㅅ→감사). Default: false */
+    jamoAbbrev?: boolean;
+    /** Normalize common dialect stems (워디→어디). Default: false */
+    dialect?: boolean;
+}
+export declare function normalizeText(text: string, opts?: NormalizeOptions): string;
+export interface Segment {
+    text: string;
+    /** Character offset of this segment in the original text */
+    offset: number;
+}
+export declare function splitSentences(text: string): Segment[];

package/dist/normalize.js

@@ -0,0 +1,71 @@
+/**
+ * Text normalization utilities for Garu Korean morphological analyzer.
+ * Pre-processes informal/non-standard Korean text before analysis.
+ */
+const COMPAT_JAMO_RE = /([\u3131-\u3163])\1{2,}/g;
+const JAMO_ABBREV_MAP = {
+    'ㄱㅅ': '감사',
+    'ㅊㅋ': '축하',
+    'ㄴㄴ': '노노',
+    'ㅂㅂ': '바이바이',
+    'ㅈㅅ': '죄송',
+    'ㄱㅊ': '괜찮',
+    'ㅇㅈ': '인정',
+    'ㅎㅇ': '하이',
+    'ㄹㅇ': '리얼',
+    'ㅇㅋ': '오케이',
+    'ㅁㄹ': '모름',
+    'ㅅㄱ': '수고',
+    'ㅊㅊ': '추천',
+    'ㅍㅌ': '파이팅',
+    'ㄷㄷ': '덜덜',
+};
+const DIALECT_MAP = {
+    '워디': '어디',
+    '와케': '왜이렇게',
+    '우예': '어떻게',
+    '긍게': '그러니까',
+    '워메': '어머',
+    '가유': '가요',
+    '허유': '해요',
+};
+export function normalizeText(text, opts = {}) {
+    const { jamo = true, jamoAbbrev = false, dialect = false } = opts;
+    let result = text;
+    if (jamo) {
+        result = result.replace(COMPAT_JAMO_RE, '$1$1');
+    }
+    if (jamoAbbrev) {
+        for (const [abbrev, expansion] of Object.entries(JAMO_ABBREV_MAP)) {
+            const re = new RegExp(`(?<![가-힣ㄱ-ㅣ])${abbrev}(?![가-힣ㄱ-ㅣ])`, 'g');
+            result = result.replace(re, expansion);
+        }
+    }
+    if (dialect) {
+        for (const [dialectForm, standard] of Object.entries(DIALECT_MAP)) {
+            result = result.replace(new RegExp(dialectForm, 'g'), standard);
+        }
+    }
+    return result;
+}
+export function splitSentences(text) {
+    const segments = [];
+    const re = /([.!?…]+)\s+(?=[가-힣A-Z"'([{])/g;
+    let lastCharOffset = 0;
+    let lastByteIdx = 0;
+    let match;
+    while ((match = re.exec(text)) !== null) {
+        const segText = text.slice(lastByteIdx, match.index + match[1].length).trim();
+        if (segText.length > 0) {
+            segments.push({ text: segText, offset: lastCharOffset });
+        }
+        const newStart = match.index + match[0].length;
+        lastCharOffset += [...text.slice(lastByteIdx, newStart)].length;
+        lastByteIdx = newStart;
+    }
+    const last = text.slice(lastByteIdx).trim();
+    if (last.length > 0) {
+        segments.push({ text: last, offset: lastCharOffset });
+    }
+    return segments.length > 0 ? segments : [{ text: text.trim(), offset: 0 }];
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "garu-ko",
-  "version": "0.3.1",
+  "version": "0.3.3",
   "description": "Ultra-lightweight Korean morphological analyzer for the web (1.7MB model, WASM 93KB, F1 95.3%)",
   "type": "module",
   "main": "dist/index.js",
@@ -12,9 +12,10 @@
     "README.md"
   ],
   "scripts": {
-    "build:wasm": "wasm-pack build ../crates/garu-wasm --target web --out-dir ../../js/pkg && echo '' > pkg/.gitignore",
+    "build:wasm": "wasm-pack build ../crates/garu-wasm --target web --out-dir ../../js/pkg && rm -f pkg/.gitignore",
     "build:js": "tsc",
     "build": "npm run build:wasm && npm run build:js",
+    "prepublishOnly": "rm -f pkg/.gitignore",
     "test": "vitest run"
   },
   "keywords": [

package/pkg/garu_wasm.js

@@ -82,7 +82,7 @@ export class GaruWasm {
     }
 }
 if (Symbol.dispose) GaruWasm.prototype[Symbol.dispose] = GaruWasm.prototype.free;
-import * as import1 from "./snippets/garu-core-d0bd202007f7afa3/inline0.js"
+import * as import1 from "./snippets/garu-core-8782e7e2eee661b3/inline0.js"
 
 function __wbg_get_imports() {
     const import0 = {
@@ -143,7 +143,7 @@ function __wbg_get_imports() {
     return {
         __proto__: null,
         "./garu_wasm_bg.js": import0,
-        "./snippets/garu-core-d0bd202007f7afa3/inline0.js": import1,
+        "./snippets/garu-core-8782e7e2eee661b3/inline0.js": import1,
     };
 }
 

package/pkg/package.json

@@ -1,7 +1,7 @@
 {
   "name": "garu-wasm",
   "type": "module",
-  "version": "0.2.10",
+  "version": "0.3.1",
   "files": [
     "garu_wasm_bg.wasm",
     "garu_wasm.js",

package/pkg/snippets/garu-core-8782e7e2eee661b3/inline0.js

@@ -0,0 +1 @@
+export function performance_now() { return performance.now(); }