Merge branch 'languagedetector-lingua' into 'main'

Add lingua backend for language detection

Closes #250

See merge request soapbox-pub/ditto!610

deno.json

@@ -19,6 +19,7 @@
     "setup:kind0": "deno run -A --env-file --deny-read=.env scripts/setup-kind0.ts",
     "stats:recompute": "deno run -A --env-file --deny-read=.env scripts/stats-recompute.ts",
     "soapbox": "curl -O https://dl.soapbox.pub/main/soapbox.zip && mkdir -p public && mv soapbox.zip public/ && cd public/ && unzip -o soapbox.zip && rm soapbox.zip",
+    "lingua": "mkdir -p data/lingua && cd data/lingua && curl -# -o lingua-wasm.zip -L https://github.com/xyzshantaram/lingua-wasm/releases/latest/download/lingua-wasm.zip && unzip -o lingua-wasm.zip && rm lingua-wasm.zip",
     "trends": "deno run -A --env-file --deny-read=.env scripts/trends.ts",
     "clean:deps": "deno cache --reload src/app.ts",
     "db:populate-search": "deno run -A --env-file --deny-read=.env scripts/db-populate-search.ts",

src/config.ts

@@ -303,6 +303,9 @@ class Conf {
   static get translationProvider(): string | undefined {
     return Deno.env.get('TRANSLATION_PROVIDER');
   }
+  static get languageDetector(): string {
+    return Deno.env.get('DITTO_LANG_DETECTOR') || 'lande';
+  }
   /** DeepL URL endpoint. */
   static get deeplBaseUrl(): string | undefined {
     return Deno.env.get('DEEPL_BASE_URL');

src/pipeline.ts

@@ -234,7 +234,7 @@ async function parseMetadata(event: NostrEvent, signal: AbortSignal): Promise<vo
 async function setLanguage(event: NostrEvent): Promise<void> {
   if (event.kind !== 1) return;
 
-  const language = detectLanguage(event.content, 0.90);
+  const language = await detectLanguage(event.content, 0.90);
   if (!language) return;
 
   const kysely = await Storages.kysely();

src/utils/language.test.ts

@@ -1,28 +1,52 @@
 import { detectLanguage } from '@/utils/language.ts';
-import { assertEquals } from '@std/assert';
+import { assertEquals, assertNotEquals } from '@std/assert';
+import { Conf } from '@/config.ts';
 
-Deno.test('Detect English language', () => {
-  assertEquals(detectLanguage(``, 0.90), undefined);
-  assertEquals(detectLanguage(`Good morning my fellow friends`, 0.90), 'en');
-  assertEquals(
-    detectLanguage(
-      `Would you listen to Michael Jackson's songs?\n\nnostr:nevent1qvzqqqqqqypzqprpljlvcnpnw3pejvkkhrc3y6wvmd7vjuad0fg2ud3dky66gaxaqyvhwumn8ghj7cm0vfexzen4d4sjucm0d5hhyetvv9usqg8htx8xcjq7ffrzxu7nrhlr8vljcv6gpmet0auy87mpj6djxk4myqha02kp`,
-      0.90,
-    ),
-    'en',
-  );
-  assertEquals(
-    detectLanguage(
-      `https://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uWould you listen to Michael Jackson's songs?\n\nnostr:nevent1qvzqqqqqqypzqprpljlvcnpnw3pejvkkhrc3y6wvmd7vjuad0fg2ud3dky66gaxaqyvhwumn8ghj7cm0vfexzen4d4sjucm0d5hhyetvv9usqg8htx8xcjq7ffrzxu7nrhlr8vljcv6gpmet0auy87mpj6djxk4myqha02kp`,
-      0.90,
-    ),
-    'en',
-  );
-  assertEquals(
-    detectLanguage(
-      `https://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_u 😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎    https://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_u Would you listen to Michael Jackson's songs?\n\nnostr:nevent1qvzqqqqqqypzqprpljlvcnpnw3pejvkkhrc3y6wvmd7vjuad0fg2ud3dky66gaxaqyvhwumn8ghj7cm0vfexzen4d4sjucm0d5hhyetvv9usqg8htx8xcjq7ffrzxu7nrhlr8vljcv6gpmet0auy87mpj6djxk4myqha02kp`,
-      0.90,
-    ),
-    'en',
-  );
+Deno.test('Tests for language detection', async (t) => {
+  await t.step('Empty string should return undefined', async () => {
+    assertEquals(await detectLanguage(``, 0.90), undefined);
+  });
+
+  await t.step('Regular English string should be detected', async () => {
+    assertEquals(await detectLanguage(`Good morning my fellow friends`, 0.90), 'en');
+  });
+
+  await t.step('nostr event id should be ignored', async () => {
+    assertEquals(
+      await detectLanguage(
+        `Would you listen to Michael Jackson's songs?\n\nnostr:nevent1qvzqqqqqqypzqprpljlvcnpnw3pejvkkhrc3y6wvmd7vjuad0fg2ud3dky66gaxaqyvhwumn8ghj7cm0vfexzen4d4sjucm0d5hhyetvv9usqg8htx8xcjq7ffrzxu7nrhlr8vljcv6gpmet0auy87mpj6djxk4myqha02kp`,
+        0.90,
+      ),
+      'en',
+    );
+  });
+
+  await t.step('URLs should be ignored', async () => {
+    assertEquals(
+      await detectLanguage(
+        `https://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uWould you listen to Michael Jackson's songs?\n\nnostr:nevent1qvzqqqqqqypzqprpljlvcnpnw3pejvkkhrc3y6wvmd7vjuad0fg2ud3dky66gaxaqyvhwumn8ghj7cm0vfexzen4d4sjucm0d5hhyetvv9usqg8htx8xcjq7ffrzxu7nrhlr8vljcv6gpmet0auy87mpj6djxk4myqha02kp`,
+        0.90,
+      ),
+      'en',
+    );
+  });
+
+  await t.step('Emoji should be ignored', async () => {
+    assertEquals(
+      await detectLanguage(
+        `https://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_u 😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎😂💯♡⌨︎    https://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_uhttps://youtu.be/FxppefYTA2I?si=grgEpbEhFu_-3V_u Would you listen to Michael Jackson's songs?\n\nnostr:nevent1qvzqqqqqqypzqprpljlvcnpnw3pejvkkhrc3y6wvmd7vjuad0fg2ud3dky66gaxaqyvhwumn8ghj7cm0vfexzen4d4sjucm0d5hhyetvv9usqg8htx8xcjq7ffrzxu7nrhlr8vljcv6gpmet0auy87mpj6djxk4myqha02kp`,
+        0.90,
+      ),
+      'en',
+    );
+  });
+
+  await t.step('The horrific problem sentence', async () => {
+    const tester = Conf.languageDetector === 'lingua' ? assertEquals : assertNotEquals;
+    tester(await detectLanguage(`It may die when I die, and that's okay. It's my earnings.`, 0.90), 'en');
+  });
+  //
+  await t.step('Should detect Hindi sentences', async () => {
+    assertEquals(await detectLanguage(`मै डिट्टो की नयी अनुवाद सुविधा को आज़मा रहा हूँ`, 0.80), 'hi');
+  });
 });

src/utils/language.ts

@@ -1,34 +1,98 @@
 import ISO6391, { type LanguageCode } from 'iso-639-1';
 import lande from 'lande';
 import linkify from 'linkifyjs';
+import { Conf } from '@/config.ts';
+
+declare class LinguaDetector {
+  private worker: Worker;
+  private pending: Map<string, PromiseWithResolvers<Map<string, number>>>;
+  destroyed: boolean;
+
+  /**
+   * Instantiate a new LanguageDetector.
+   * @param workerPath Path to the worker file. If not supplied, it defaults to using the lingua-wasm build present with this package.
+   * The worker is simply posted messages with a detection id and the string to detect. It must then respond with the same detection id and the detected language (or undefined.)
+   */
+  constructor(workerPath?: string);
+
+  /**
+   * Detect the language of a string.
+   * @param str The string to detect for.
+   * @returns A Promise that resolves to an ISO-639-3 language code.
+   */
+  detect(str: string): Promise<string>;
+
+  /**
+   * Checks if the LanguageDetector has been destroyed.
+   * @throws Error if the detector has been destroyed.
+   */
+  private checkDestroyed(): void;
+
+  /**
+   * Destroys the LanguageDetector instance and terminates the associated Worker.
+   */
+  destroy(): void;
+}
+
+const toLanguageCode = (iso6393: string) => {
+  const locale = new Intl.Locale(iso6393);
+  if (ISO6391.validate(locale.language)) return locale.language as LanguageCode;
+};
+
+function detectWithLande(sanitized: string, threshold: number) {
+  const [topResult] = lande(sanitized);
+  if (topResult) {
+    const [iso6393, confidence] = topResult;
+
+    if (confidence >= threshold) {
+      return toLanguageCode(iso6393);
+    }
+  }
+}
+
+const linguaDirUrl = (file: string) => {
+  const normalized = new URL('../../data/lingua/' + file, import.meta.url);
+  return normalized.toString();
+};
+
+let linguaInstance: LinguaDetector | undefined = undefined;
+
+async function detectWithLingua(text: string) {
+  try {
+    if (!linguaInstance) {
+      const { LanguageDetector } = await import(linguaDirUrl('mod.ts'));
+      linguaInstance = new LanguageDetector(linguaDirUrl('src/worker.js')) as unknown as LinguaDetector;
+      globalThis.addEventListener('unload', () => linguaInstance?.destroy());
+    }
+    const result = await linguaInstance.detect(text);
+    return toLanguageCode(result);
+  } catch (e) {
+    console.error(e);
+    return null;
+  }
+}
 
 linkify.registerCustomProtocol('nostr', true);
 
 /** Returns the detected language if the confidence is greater or equal than 'minConfidence'
  * 'minConfidence' must be a number between 0 and 1, such as 0.95
  */
-export function detectLanguage(text: string, minConfidence: number): LanguageCode | undefined {
+export async function detectLanguage(text: string, threshold: number): Promise<LanguageCode | undefined> {
   // It's better to remove the emojis first
-  const sanitizedText = linkify.tokenize(
+  const sanitized = linkify.tokenize(
     text
       .replaceAll(/\p{Extended_Pictographic}/gu, '')
       .replaceAll(/[\s\uFEFF\u00A0\u200B-\u200D\u{0FE0E}]+/gu, ' '),
   ).reduce((acc, { t, v }) => t === 'text' ? acc + v : acc, '').trim();
 
-  if (sanitizedText.length < 10) { // heuristics
+  if (sanitized.length < 10) { // heuristics
     return;
   }
 
-  const [topResult] = lande(
-    sanitizedText,
-  );
-  if (topResult) {
-    const [iso6393, confidence] = topResult;
-    const locale = new Intl.Locale(iso6393);
-
-    if (confidence >= minConfidence && ISO6391.validate(locale.language)) {
-      return locale.language as LanguageCode;
-    }
+  if (Conf.languageDetector === 'lingua') {
+    const detected = await detectWithLingua(sanitized);
+    if (detected) return detected;
   }
-  return;
+
+  return detectWithLande(sanitized, threshold);
 }