Merge branch 'improve-lande-use' into 'main'

Improve setLanguage() function for better language detection

See merge request soapbox-pub/ditto!543

src/pipeline.ts

@@ -1,8 +1,6 @@
 import { NKinds, NostrEvent, NSchema as n } from '@nostrify/nostrify';
 import { Stickynotes } from '@soapbox/stickynotes';
-import ISO6391 from 'iso-639-1';
 import { Kysely, sql } from 'kysely';
-import lande from 'lande';
 import { LRUCache } from 'lru-cache';
 import { z } from 'zod';
 
@@ -22,6 +20,7 @@ import { nip05Cache } from '@/utils/nip05.ts';
 import { purifyEvent } from '@/utils/purify.ts';
 import { updateStats } from '@/utils/stats.ts';
 import { getTagSet } from '@/utils/tags.ts';
+import { detectLanguage } from '@/utils/language.ts';
 
 const console = new Stickynotes('ditto:pipeline');
 
@@ -200,23 +199,19 @@ async function parseMetadata(event: NostrEvent, signal: AbortSignal): Promise<vo
 
 /** Update the event in the database and set its language. */
 async function setLanguage(event: NostrEvent): Promise<void> {
-  const [topResult] = lande(event.content);
+  if (event.kind !== 1) return;
 
-  if (topResult) {
-    const [iso6393, confidence] = topResult;
-    const locale = new Intl.Locale(iso6393);
+  const language = detectLanguage(event.content, 0.90);
+  if (!language) return;
 
-    if (confidence >= 0.95 && ISO6391.validate(locale.language)) {
-      const kysely = await Storages.kysely();
-      try {
-        await kysely.updateTable('nostr_events')
-          .set('language', locale.language)
-          .where('id', '=', event.id)
-          .execute();
-      } catch {
-        // do nothing
-      }
-    }
+  const kysely = await Storages.kysely();
+  try {
+    await kysely.updateTable('nostr_events')
+      .set('language', language)
+      .where('id', '=', event.id)
+      .execute();
+  } catch {
+    // do nothing
   }
 }
 

src/utils/language.test.ts

@@ -0,0 +1,28 @@
+import { detectLanguage } from '@/utils/language.ts';
+import { assertEquals } from '@std/assert';
+
+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',
+  );
+});

src/utils/language.ts

@@ -0,0 +1,34 @@
+import ISO6391, { type LanguageCode } from 'iso-639-1';
+import lande from 'lande';
+import linkify from 'linkifyjs';
+
+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 {
+  // It's better to remove the emojis first
+  const sanitizedText = 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
+    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;
+    }
+  }
+  return;
+}