added multilang support

2026-06-25 18:11:18 -05:00
parent c63a35dd94
commit d047e896be
11 changed files with 436 additions and 196 deletions
@@ -1,12 +1,11 @@
 {
  "manifest_version": 3,
  "name": "InTruth",
-  "version": "1.1.0",
+  "version": "1.1.9",
  "description": "Real-time fact-checking of political speech and debates",

  "permissions": [
    "activeTab",
-    "scripting",
    "storage",
    "tabCapture",
    "offscreen"
@@ -1,17 +1,13 @@
-// service-worker.js --> now combined with fact-checker.js and claim-detector.js for import conflicts
-// transcription runs in content script via web speech API;
-// responsibile for starting / stopping content script, receiving transcripts,
-// and routing them to claim detection
-// 5.31.2026 -- serper call before claude call for more accurate verdicts
-// 6.12.2026 -- switch to deepgram; too many conflicts w/ webaudio
-
+// service-worker.js
 let ANTHROPIC_KEY = '';
 const SERPER_KEY = '';
+let TRANSCRIPT_LANGUAGE = 'en';

 async function loadKeys() {
  return new Promise(resolve => {
-    chrome.storage.local.get(['anthropicKey'], (data) => {
+    chrome.storage.local.get(['anthropicKey', 'transcriptLanguage'], (data) => {
      ANTHROPIC_KEY = data.anthropicKey || '';
+      TRANSCRIPT_LANGUAGE = data.transcriptLanguage || 'en';
      resolve();
    });
  });
@@ -19,24 +15,50 @@ async function loadKeys() {

 const EVALUATE_PROMPT = ``;

-// ── Speaker parsing (mirrors overlay.js) ─────────────────────────────────────
+
+const GROUNDED_PROMPT = ``;
+
+
+const SPEAKER_PARSE_NOISE = new Set(['debate','presidential','vp','vice','2024','2023','2022','2021','2020','2019','2016','surrounded','tonight','live','full','official']);
 
 function parseSpeakersFromTitle(title) {
  if (!title) return [];
-  const roleMatch = title.match(/(\d+)\s+([a-z]+(?:\s+[a-z]+)?)\s+(?:vs?\.?|versus)\s+(\d+)\s+([a-z]+(?:\s+[a-z]+)?)/i);
+  const clean = title.split('|')[0].trim();
+ 
+  // 'N role vs N role' — e.g. '1 Liberal vs 20 Conservatives'
+  const roleMatch = clean.match(/(\d+)\s+([a-z]+(?:\s+[a-z]+)?)\s+(?:vs?\.?|versus)\s+(\d+)\s+([a-z]+(?:\s+[a-z]+)?)/i);
  if (roleMatch) {
    const cap = s => s.charAt(0).toUpperCase() + s.slice(1);
    return [cap(roleMatch[2]), cap(roleMatch[4])];
  }
-  const nameMatch = title.match(/([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)\s+(?:and|vs\.?|versus|&)\s+([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)/);
-  if (nameMatch) {
-    const clean = name => name.trim().split(' ').pop();
-    return [clean(nameMatch[1]), clean(nameMatch[2])];
+ 
+  // 'Name vs N Description' — second side starts with digit, e.g. 'Dean Withers vs 20 MAGA Women'
+  const nameVsGroupMatch = clean.match(/([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)\s+(?:vs?\.?|versus)\s+(\d+)\s+(.+)/i);
+  if (nameVsGroupMatch) {
+    const name = nameVsGroupMatch[1].trim().split(' ').pop();
+    const groupWords = nameVsGroupMatch[3].trim().split(/\s+/);
+    const group = groupWords.filter(w => !SPEAKER_PARSE_NOISE.has(w.toLowerCase())).pop() || groupWords.pop();
+    return [name, group];
  }
+ 
+  // split on vs/and — take last non-noise capitalized word from each side
+  const vsSplit = clean.split(/\s+(?:vs?\.?|versus|and|&)\s+/i);
+  if (vsSplit.length >= 2) {
+    const lastName = part => {
+      const words = part.trim().split(/\s+/);
+      for (let i = words.length - 1; i >= 0; i--) {
+        if (/^[A-Z]/.test(words[i]) && !SPEAKER_PARSE_NOISE.has(words[i].toLowerCase())) return words[i];
+      }
+      return null;
+    };
+    const a = lastName(vsSplit[0]);
+    const b = lastName(vsSplit[1]);
+    if (a && b) return [a, b];
+  }
+ 
  return [];
 }
 
-// ── Serper ────────────────────────────────────────────────────────────────────
 
 const BLOCKED_DOMAINS = [
  'reddit.com', 'facebook.com', 'twitter.com', 'x.com',
@@ -48,30 +70,71 @@ const BLOCKED_DOMAINS = [
  'thefederalist.com', 'motherjones.com', 'nationalreview.com',
  'democrats-appropriations.house.gov', 'waysandmeans.house.gov',
  'bostonkravmaga.com',
+  'israelpolicyforum.org',
 ];
 
+const LANGUAGE_LOCALE = {
+  en: { gl: 'us', hl: 'en' },
+  es: { gl: 'es', hl: 'es' },
+  fr: { gl: 'fr', hl: 'fr' },
+  de: { gl: 'de', hl: 'de' },
+  it: { gl: 'it', hl: 'it' },
+  pt: { gl: 'br', hl: 'pt' },
+  nl: { gl: 'nl', hl: 'nl' },
+  hi: { gl: 'in', hl: 'hi' },
+  ja: { gl: 'jp', hl: 'ja' },
+  zh: { gl: 'cn', hl: 'zh-cn' },
+  ar: { gl: 'sa', hl: 'ar' },
+  ko: { gl: 'kr', hl: 'ko' },
+  ru: { gl: 'ru', hl: 'ru' },
+  pl: { gl: 'pl', hl: 'pl' },
+  sv: { gl: 'se', hl: 'sv' },
+  tr: { gl: 'tr', hl: 'tr' },
+};
+
 async function searchWeb(query, retries = 2) {
  try {
    const res = await fetch('https://google.serper.dev/search', {
      method: 'POST',
      headers: { 'Content-Type': 'application/json', 'X-API-KEY': SERPER_KEY },
-      body: JSON.stringify({ q: query, num: 6 }),
+      body: JSON.stringify({ q: query, num: 6, ...(LANGUAGE_LOCALE[TRANSCRIPT_LANGUAGE] || LANGUAGE_LOCALE.en) }),
    });
    const data = await res.json();
-    return (data.organic ?? [])
-      .map(r => r.link)
-      .filter(url => url && !BLOCKED_DOMAINS.some(d => url.includes(d)))
-      .slice(0, 3);
+ 
+    const organic = (data.organic ?? [])
+      .filter(r => r.link && !BLOCKED_DOMAINS.some(d => r.link.includes(d)))
+      .slice(0, 3)
+      .map(r => ({ url: r.link, title: r.title || '', snippet: r.snippet || '', date: r.date || '' }));
+ 
+    // answerBox — Google's direct factual answer, highest quality signal
+    const answerBox = data.answerBox
+      ? {
+          answer: data.answerBox.answer || data.answerBox.snippet || '',
+          title:  data.answerBox.title  || '',
+          url:    data.answerBox.link   || '',
+        }
+      : null;
+ 
+    // knowledgeGraph — structured entity data
+    const knowledgeGraph = data.knowledgeGraph
+      ? {
+          description: data.knowledgeGraph.description || '',
+          title:       data.knowledgeGraph.title       || '',
+        }
+      : null;
+ 
+    return { organic, answerBox, knowledgeGraph };
  } catch (err) {
    if (retries > 0) {
      await new Promise(r => setTimeout(r, 500));
      return searchWeb(query, retries - 1);
    }
    console.error('[serper] error:', err);
-    return [];
+    return { organic: [], answerBox: null, knowledgeGraph: null };
  }
 }

+ 
 // ── Claude ────────────────────────────────────────────────────────────────────
 
 async function callClaude(userMessage, systemPrompt) {
@@ -140,12 +203,12 @@ function extractLexical(text) {
 function buildLexicalSummary(f) {
  const r = f.rates || f;
  const notes = [];
-  if (r.hedging > 8)       notes.push(`hedging language (${r.hedging}%)`);
-  if (r.certainty > 8)     notes.push(`certainty markers (${r.certainty}%)`);
-  if (r.filler > 8)        notes.push(`filler words (${r.filler}%)`);
-  if (r.emotional > 8)     notes.push(`emotional language (${r.emotional}%)`);
-  if (r.exclusive > 8)     notes.push(`qualifying words (${r.exclusive}%)`);
-  if (r.firstPersonSg > 8) notes.push(`first-person singular (${r.firstPersonSg}%)`);
+  if (r.hedging > 5)       notes.push(`hedging language (${r.hedging}%)`);
+  if (r.certainty > 5)     notes.push(`certainty markers (${r.certainty}%)`);
+  if (r.filler > 5)        notes.push(`filler words (${r.filler}%)`);
+  if (r.emotional > 5)     notes.push(`emotional language (${r.emotional}%)`);
+  if (r.exclusive > 5)     notes.push(`qualifying words (${r.exclusive}%)`);
+  if (r.firstPersonSg > 5) notes.push(`first-person singular (${r.firstPersonSg}%)`);
  if (f.wordsPerSecond) {
    const pace = f.wordsPerSecond > 3.5 ? 'fast' : f.wordsPerSecond < 2 ? 'slow' : 'moderate';
    notes.push(`speech rate ${f.wordsPerSecond} w/s (${pace})`);
@@ -207,7 +270,7 @@ const WINDOW_KEEP = 15;
 // Each entry: { text, speakerId, speakerName }
 let sentenceWindow  = [];
 let sentenceCount   = 0;
-let windowLexical   = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0 };
+let windowLexical   = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0, _sentenceCount: 0 };
 let windowStartTime = null;
 let pageTitle       = '';
 let pageDate        = '';
@@ -218,7 +281,7 @@ let confirmedSpeakers = new Set(); // IDs that have been confirmed by user
 function resetWindow() {
  sentenceWindow   = [];
  sentenceCount    = 0;
-  windowLexical    = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0 };
+  windowLexical    = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0, _sentenceCount: 0 };
  windowStartTime  = null;
  currentSpeakerId  = null;
  lastSpeakerId     = null;
@@ -246,8 +309,16 @@ async function onNewSentence(text, speakerId) {
      : null;
    const flushDominantSpeaker = flushDominantId !== null ? (speakerIdToName[flushDominantId] || null) : null;
    const flushLexSnapshot = JSON.parse(JSON.stringify(windowLexical));
+    const fsc = flushLexSnapshot._sentenceCount || 1;
+    const flr = flushLexSnapshot.rates;
+    flr.hedging       = Math.round(flr.hedging       / fsc);
+    flr.certainty     = Math.round(flr.certainty     / fsc);
+    flr.filler        = Math.round(flr.filler        / fsc);
+    flr.emotional     = Math.round(flr.emotional     / fsc);
+    flr.exclusive     = Math.round(flr.exclusive     / fsc);
+    flr.firstPersonSg = Math.round(flr.firstPersonSg / fsc);
    const flushLexSummary  = buildLexicalSummary(flushLexSnapshot);
-    windowLexical   = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0 };
+    windowLexical   = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0, _sentenceCount: 0 };
    windowStartTime = null;
    await evaluateClaims(flushText, pageTitle, flushLexSummary, flushLexSnapshot, flushDominantSpeaker, flushDominantId);
  }
@@ -264,16 +335,17 @@ async function onNewSentence(text, speakerId) {
 
  if (!windowStartTime) windowStartTime = Date.now();
 
-  // accumulate lexical
+  // accumulate lexical — running sum, divide by sentence count at snapshot time
  const f = extractLexical(text);
  const r = f.rates, wr = windowLexical.rates;
-  wr.hedging       = Math.round((wr.hedging       + r.hedging)       / 2);
-  wr.certainty     = Math.round((wr.certainty     + r.certainty)     / 2);
-  wr.filler        = Math.round((wr.filler        + r.filler)        / 2);
-  wr.emotional     = Math.round((wr.emotional     + r.emotional)     / 2);
-  wr.exclusive     = Math.round((wr.exclusive     + r.exclusive)     / 2);
-  wr.firstPersonSg = Math.round((wr.firstPersonSg + r.firstPersonSg) / 2);
+  wr.hedging       += r.hedging;
+  wr.certainty     += r.certainty;
+  wr.filler        += r.filler;
+  wr.emotional     += r.emotional;
+  wr.exclusive     += r.exclusive;
+  wr.firstPersonSg += r.firstPersonSg;
  windowLexical.wordCount += f.wordCount;
+  windowLexical._sentenceCount = (windowLexical._sentenceCount || 0) + 1;
 
  if (sentenceCount % WINDOW_SIZE === 0) {
    const contextText = sentenceWindow.map(s => s.text).join(' ');
@@ -301,10 +373,19 @@ async function onNewSentence(text, speakerId) {
    windowStartTime = null;
 
    const lexicalSnapshot = JSON.parse(JSON.stringify(windowLexical));
+    // average the accumulated sums now that we have the full window
+    const sc = lexicalSnapshot._sentenceCount || 1;
+    const lr = lexicalSnapshot.rates;
+    lr.hedging       = Math.round(lr.hedging       / sc);
+    lr.certainty     = Math.round(lr.certainty     / sc);
+    lr.filler        = Math.round(lr.filler        / sc);
+    lr.emotional     = Math.round(lr.emotional     / sc);
+    lr.exclusive     = Math.round(lr.exclusive     / sc);
+    lr.firstPersonSg = Math.round(lr.firstPersonSg / sc);
    const lexicalSummary  = buildLexicalSummary(lexicalSnapshot);
 
    // reset for next window
-    windowLexical   = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0 };
+    windowLexical   = { rates: { hedging: 0, certainty: 0, filler: 0, emotional: 0, exclusive: 0, firstPersonSg: 0 }, wordsPerSecond: null, wordCount: 0, _sentenceCount: 0 };
    windowStartTime = null;
 
    try {
@@ -333,9 +414,13 @@ async function evaluateClaims(contextText, title, lexicalSummary, lexicalSnapsho
        `\n- NEVER output "Speaker N" or any [Speaker N] format in any field.`
      : `\nIdentify speakers using first-person language, policy content, and speech patterns. Never output "Speaker N".`;
 
-    const titleContext = title
-      ? `Video: "${title}"${dateContext}${speakerLegend}\n\nEvaluate claims as they were made at the time of this recording. Do not apply knowledge of events after this date.\n\n`
+    const languageInstruction = TRANSCRIPT_LANGUAGE && TRANSCRIPT_LANGUAGE !== 'en'
+      ? `\nLANGUAGE REQUIREMENT: You MUST write the "claim" and "explanation" fields in ${TRANSCRIPT_LANGUAGE}. This is mandatory regardless of what language your sources are in. Only the verdict values (TRUE, FALSE, etc) stay in English.`
      : '';
+ 
+    const titleContext = title
+      ? `Video: "${title}"${dateContext}${speakerLegend}\n\nEvaluate claims as they were made at the time of this recording. Do not apply knowledge of events after this date.${languageInstruction}\n\n`
+      : languageInstruction ? `${languageInstruction}\n\n` : '';
    const lexicalContext = lexicalSummary ? `\n\nLexical analysis: ${lexicalSummary}` : '';
 
    // already-checked claims list for Claude
@@ -348,15 +433,19 @@ async function evaluateClaims(contextText, title, lexicalSummary, lexicalSnapsho
      ? `\n\nClaims already fact-checked this session — do NOT re-evaluate these or close variants:\n- ${checkedList}\n`
      : '';
 
+    // fast Claude call — Serper searches fire immediately after on the returned claims
    const raw     = await callClaude(
      `${titleContext}Transcript: "${contextText}"${alreadyChecked}${lexicalContext}`,
      EVALUATE_PROMPT
    );
    const results = parseArray(raw);
-    const valid   = results.filter(r => r.claim && r.verdict && !isDuplicate(r.claim));
+    const valid   = results.filter(r => r.claim && r.verdict && r.verdict !== 'UNVERIFIABLE' && !isDuplicate(r.claim));
 
    if (!valid.length) return;
 
+    // kick off per-claim Serper searches in parallel with sending fast cards to overlay
+    const claimSearchPromises = valid.map(r => searchWeb(r.claim));
+ 
    if (activeTabId) {
      chrome.tabs.sendMessage(activeTabId, {
        type: 'NEW_VERDICT',
@@ -365,54 +454,79 @@ async function evaluateClaims(contextText, title, lexicalSummary, lexicalSnapsho
          sources:          [],
          pending:          true,
          lexical:          lexicalSnapshot,
-          dominantSpeakerId, // raw Deepgram ID — overlay resolves to name at render time
+          dominantSpeakerId,
          speaker:          dominantSpeaker || (r.speaker && !r.speaker.match(/^Speaker\s*\d+$/i) ? r.speaker : null),
        })),
      }).catch(() => {});
      console.log('[pipeline] fast verdicts sent:', valid.length, '| speaker:', dominantSpeaker);
    }
 
-    groundAndUpdate(contextText, valid, title, lexicalSummary, lexicalSnapshot, dominantSpeaker, dominantSpeakerId);
+    groundAndUpdate(contextText, valid, title, lexicalSummary, lexicalSnapshot, dominantSpeaker, dominantSpeakerId, claimSearchPromises);
 
  } catch (err) {
    console.error('[pipeline] error:', err);
  }
 }
 
-async function groundAndUpdate(contextText, fastResults, title, lexicalSummary, lexicalSnapshot, dominantSpeaker, dominantSpeakerId) {
+async function groundAndUpdate(contextText, fastResults, title, lexicalSummary, lexicalSnapshot, dominantSpeaker, dominantSpeakerId, claimSearchPromises = null) {
  try {
    const dateCtx      = pageDate ? `\nDate: ${pageDate}` : '';
-    const titleContext = title
-      ? `Video: "${title}"${dateCtx}\nEvaluate claims as they were made at the time of this recording. Web search results may include articles published after the debate date — ignore any information that was not publicly known at the time of the debate.\n\n`
+    const languageInstruction = TRANSCRIPT_LANGUAGE && TRANSCRIPT_LANGUAGE !== 'en'
+      ? `\nLANGUAGE REQUIREMENT: You MUST write the "claim" and "explanation" fields in ${TRANSCRIPT_LANGUAGE}. This is mandatory regardless of what language your sources are in. Only the verdict values (TRUE, FALSE, etc) stay in English.`
      : '';
+ 
+    const titleContext = title
+      ? `Video: "${title}"${dateCtx}\nEvaluate claims as they were made at the time of this recording. Web search results may include articles published after the debate date — ignore any information that was not publicly known at the time of the debate.${languageInstruction}\n\n`
+      : languageInstruction ? `${languageInstruction}\n\n` : '';
    const lexicalContext = lexicalSummary ? `\n\nLexical analysis: ${lexicalSummary}` : '';
 
-    const groundedAll = await Promise.all(fastResults.map(async (fastResult) => {
+    const groundedAll = await Promise.all(fastResults.map(async (fastResult, i) => {
      try {
-        const urls = await searchWeb(fastResult.claim);
-        if (!urls.length) return null;
+        const searchData = claimSearchPromises
+          ? await claimSearchPromises[i]
+          : await searchWeb(fastResult.claim);
+        if (!searchData.organic?.length && !searchData.answerBox && !searchData.knowledgeGraph) {
+          // no search results — finalize with fast verdict so card doesn't hang as pending
+          const resolvedSpeaker = dominantSpeaker || (fastResult.speaker && !fastResult.speaker.match(/^Speaker\s*\d+$/i) ? fastResult.speaker : null);
+          return { ...fastResult, sources: [], pending: false, lexical: lexicalSnapshot, speaker: resolvedSpeaker, dominantSpeakerId, _fastClaim: fastResult.claim };
+        }
+ 
+        const urls = searchData.organic.map(r => r.url);
+ 
+        // build evidence block — answerBox first (highest quality), then knowledgeGraph, then organic
+        const parts = [];
+        if (searchData.answerBox?.answer) {
+          parts.push(`[Direct Answer] ${searchData.answerBox.title ? searchData.answerBox.title + ': ' : ''}${searchData.answerBox.answer}${searchData.answerBox.url ? '\n' + searchData.answerBox.url : ''}`);
+        }
+        if (searchData.knowledgeGraph?.description) {
+          parts.push(`[Knowledge Panel] ${searchData.knowledgeGraph.title ? searchData.knowledgeGraph.title + ': ' : ''}${searchData.knowledgeGraph.description}`);
+        }
+        searchData.organic.forEach((r, idx) => {
+          const datePart = r.date ? ` (${r.date})` : '';
+          parts.push(`[${idx+1}] ${r.title}${datePart}\n${r.url}\n${r.snippet}`);
+        });
+        const evidenceBlock = parts.join('\n\n');
        const raw = await callClaude(
-          `${titleContext}Transcript: "${contextText}"\n\nEvaluate ONLY this specific claim:\n1. ${fastResult.claim}\n\nWeb search results:\n${urls.join('\n')}${lexicalContext}`,
-          EVALUATE_PROMPT
+          `${titleContext}Transcript: "${contextText}"\n\nClaim: "${fastResult.claim}"\nFast verdict: ${fastResult.verdict}\n\nWeb search evidence:\n${evidenceBlock}${lexicalContext}`,
+          GROUNDED_PROMPT
        );
-        const results = parseArray(raw);
-        const match   = results.find(r => r.claim && r.verdict);
-        if (!match) return null;
-        // re-resolve speaker at grounding time — user may have confirmed since fast pass
-        const lateResolved = dominantSpeakerId !== null && dominantSpeakerId !== undefined
-          ? speakerIdToName[dominantSpeakerId] || null
-          : null;
-        const resolvedSpeaker = lateResolved
-          || dominantSpeaker
-          || (match.speaker && !match.speaker.match(/^Speaker\s*\d+$/i) ? match.speaker : null)
-          || (fastResult.speaker && !fastResult.speaker.match(/^Speaker\s*\d+$/i) ? fastResult.speaker : null);
+        const parsed = parseArray(raw);
+        const match  = parsed.find(r => r.claim && r.verdict);
+        // drop UNVERIFIABLE from grounded pass — either it's checkable or it isn't shown
+        if (!match || match.verdict === 'UNVERIFIABLE') return null;
+        const resolvedSpeaker = dominantSpeaker
+          || (fastResult.speaker && !fastResult.speaker.match(/^Speaker\s*\d+$/i) ? fastResult.speaker : null)
+          || (match.speaker && !match.speaker.match(/^Speaker\s*\d+$/i) ? match.speaker : null);
 
-        // never downgrade TRUE to MISLEADING in grounded pass — fast verdict had no sources to nitpick
+        // code-level protection: never downgrade TRUE/SUBSTANTIALLY TRUE to MISLEADING or FALSE
+        // the grounded prompt repeatedly violates this rule by reasoning from snippets
+        // fast pass has full training knowledge; grounded pass has 1-2 sentence snippets
+        // only the grounded pass can upgrade verdicts or add SUBSTANTIALLY TRUE context
        const fastWasTrue = fastResult.verdict === 'TRUE' || fastResult.verdict === 'SUBSTANTIALLY TRUE';
-        const groundedIsMisleading = match.verdict === 'MISLEADING';
-        const finalVerdict = (fastWasTrue && groundedIsMisleading) ? fastResult.verdict : match.verdict;
+        const groundedDowngrades = match.verdict === 'MISLEADING' || match.verdict === 'FALSE';
+        const finalVerdict = (fastWasTrue && groundedDowngrades) ? fastResult.verdict : match.verdict;
 
-        return { ...match, verdict: finalVerdict, sources: urls, pending: false, lexical: lexicalSnapshot, speaker: resolvedSpeaker, dominantSpeakerId };
+        return { ...match, verdict: finalVerdict, sources: urls, pending: false, lexical: lexicalSnapshot, speaker: resolvedSpeaker, dominantSpeakerId, _fastClaim: fastResult.claim };
      } catch (err) {
        console.error('[grounded] error:', fastResult.claim.slice(0, 40), err);
        return null;
@@ -566,7 +680,7 @@ async function startFactCheck() {
    });
  });
 
-  const response = await chrome.runtime.sendMessage({ type: 'START_CAPTURE', streamId });
+  const response = await chrome.runtime.sendMessage({ type: 'START_CAPTURE', streamId, language: TRANSCRIPT_LANGUAGE });
  if (!response?.ok) throw new Error('Failed to start capture: ' + response?.error);
 
  // reset BEFORE sending START_FACTCHECK — transcripts arrive immediately after
@@ -64,19 +64,43 @@ function getSpeakerColor(name) {
 }

 // ── Speaker parsing ───────────────────────────────────────────────────────────
+const SPEAKER_PARSE_NOISE = new Set(['debate','presidential','vp','vice','2024','2023','2022','2021','2020','2019','2016','surrounded','tonight','live','full','official']);
+
 function parseSpeakersFromTitle(title) {
  if (!title) return [];
-  const roleMatch = title.match(/(\d+)\s+([a-z]+(?:\s+[a-z]+)?)\s+(?:vs?\.?|versus)\s+(\d+)\s+([a-z]+(?:\s+[a-z]+)?)/i);
+  const clean = title.split('|')[0].trim();
+
+  // 'N role vs N role' — e.g. '1 Liberal vs 20 Conservatives'
+  const roleMatch = clean.match(/(\d+)\s+([a-z]+(?:\s+[a-z]+)?)\s+(?:vs?\.?|versus)\s+(\d+)\s+([a-z]+(?:\s+[a-z]+)?)/i);
  if (roleMatch) {
    const cap = s => s.charAt(0).toUpperCase() + s.slice(1);
    return [cap(roleMatch[2]), cap(roleMatch[4])];
  }
-  // only match capitalized proper names (not lowercase words like "in", "the", etc.)
-  const nameMatch = title.match(/([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)\s+(?:and|vs\.?|versus|&)\s+([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)/);
-  if (nameMatch) {
-    const clean = name => name.trim().split(' ').pop();
-    return [clean(nameMatch[1]), clean(nameMatch[2])];
+
+  // 'Name vs N Description' — second side starts with digit, e.g. 'Dean Withers vs 20 MAGA Women'
+  const nameVsGroupMatch = clean.match(/([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)\s+(?:vs?\.?|versus)\s+(\d+)\s+(.+)/i);
+  if (nameVsGroupMatch) {
+    const name = nameVsGroupMatch[1].trim().split(' ').pop();
+    const groupWords = nameVsGroupMatch[3].trim().split(/\s+/);
+    const group = groupWords.filter(w => !SPEAKER_PARSE_NOISE.has(w.toLowerCase())).pop() || groupWords.pop();
+    return [name, group];
  }
+
+  // split on vs/and — take last non-noise capitalized word from each side
+  const vsSplit = clean.split(/\s+(?:vs?\.?|versus|and|&)\s+/i);
+  if (vsSplit.length >= 2) {
+    const lastName = part => {
+      const words = part.trim().split(/\s+/);
+      for (let i = words.length - 1; i >= 0; i--) {
+        if (/^[A-Z]/.test(words[i]) && !SPEAKER_PARSE_NOISE.has(words[i].toLowerCase())) return words[i];
+      }
+      return null;
+    };
+    const a = lastName(vsSplit[0]);
+    const b = lastName(vsSplit[1]);
+    if (a && b) return [a, b];
+  }
+
  return [];
 }

@@ -120,14 +144,15 @@ const confirmedSpeakerMap = {}; // { speakerId: 'Harris' }
 const pendingSpeakerIds   = new Set(); // IDs waiting for confirmation

 function showSpeakerBanner(speakerId, sample) {
-  if (pendingSpeakerIds.has(speakerId)) return;
-  if (speakerId in confirmedSpeakerMap) return;
+  const sid = String(speakerId);
+  if (pendingSpeakerIds.has(sid)) return;
+  if (sid in confirmedSpeakerMap) return;
  // if speakers not yet parsed from title, retry once after 1s
  if (!speakers.length) {
    setTimeout(() => showSpeakerBanner(speakerId, sample), 1000);
    return;
  }
-  pendingSpeakerIds.add(speakerId);
+  pendingSpeakerIds.add(sid);

  const banner = document.createElement('div');
  banner.className = 'rtfc-speaker-banner';
@@ -136,15 +161,15 @@ function showSpeakerBanner(speakerId, sample) {
    '<div class="rtfc-speaker-banner-sample">"' + escapeHtml(sample) + '..."</div>' +
    '<div class="rtfc-speaker-banner-buttons">' +
      speakers.map(name =>
-        '<button class="rtfc-speaker-banner-btn" data-name="' + escapeHtml(name) + '" data-id="' + speakerId + '">' + escapeHtml(name) + '</button>'
+        '<button class="rtfc-speaker-banner-btn" data-name="' + escapeHtml(name) + '" data-id="' + sid + '">' + escapeHtml(name) + '</button>'
      ).join('') +
-      '<button class="rtfc-speaker-banner-btn rtfc-speaker-banner-btn--skip" data-id="' + speakerId + '">Skip</button>' +
+      '<button class="rtfc-speaker-banner-btn rtfc-speaker-banner-btn--skip" data-id="' + sid + '">Skip</button>' +
    '</div>';

  banner.querySelectorAll('.rtfc-speaker-banner-btn').forEach(btn => {
    btn.addEventListener('click', () => {
      const name = btn.dataset.name;
-      const id   = parseInt(btn.dataset.id);
+      const id   = btn.dataset.id; // already a string — matches confirmedSpeakerMap keys
      if (name) {
        confirmedSpeakerMap[id] = name;
        chrome.runtime.sendMessage({
@@ -168,13 +193,11 @@ function showSpeakerBanner(speakerId, sample) {
 // ── Speaker confirmation state ───────────────────────────────────────────────

 function allSpeakersConfirmed() {
-  // true when every speaker seen so far has been confirmed or skipped
-  // and at least one real name has been confirmed
-  const confirmedNames = Object.values(confirmedSpeakerMap).filter(v => v !== null);
-  return confirmedNames.length >= Math.min(speakers.length, Object.keys(confirmedSpeakerMap).length)
-    && Object.keys(confirmedSpeakerMap).length > 0;
+  // true only when every speaker parsed from the title has been confirmed (or skipped) by the user
+  // Deepgram assigns IDs 0, 1, 2... in order of first appearance — matches speakers array indices
+  if (!speakers.length) return false;
+  return speakers.every((_, i) => String(i) in confirmedSpeakerMap);
 }
-
 function retryTagAllCards() {
  // retroactively tag all grounded cards once speakers are confirmed
  if (!verdictListEl) return;
@@ -468,21 +491,15 @@ function buildCard(result) {

  const lexicalRows = buildLexicalRows(result.lexical);

-  // speaker tag — only show on grounded cards AND only when all speakers confirmed
-  // this prevents wrong tags from appearing before diarization stabilizes
+  // speaker tag — only show when ALL speakers have been confirmed by user
+  // prevents wrong tags on cards detected before diarization stabilized
  let speakerTag = '';
-  if (!result.pending && allSpeakersConfirmed()) {
-    const confirmedName = (result.dominantSpeakerId !== null && result.dominantSpeakerId !== undefined)
-      ? confirmedSpeakerMap[result.dominantSpeakerId]
-      : undefined;
-    const rawSpeaker = (confirmedName !== undefined && confirmedName !== null)
-      ? confirmedName
-      : result.speaker || null;
-    const normalizedName = rawSpeaker ? normalizeSpeakerName(rawSpeaker) : null;
-    const speakerName  = (normalizedName && !normalizedName.match(/^Speaker\s*\d+$/i)) ? normalizedName : null;
-    const speakerColor = speakerName ? getSpeakerColor(speakerName) : null;
-    if (speakerColor) {
-      speakerTag = '<div class="rtfc-speaker-tag" style="background:' + speakerColor + '">' + escapeHtml(speakerName) + '</div>';
+  if (!result.pending && allSpeakersConfirmed() && result.dominantSpeakerId !== null && result.dominantSpeakerId !== undefined) {
+    const confirmedName = confirmedSpeakerMap[result.dominantSpeakerId];
+    if (confirmedName) {
+      const name = normalizeSpeakerName(confirmedName);
+      const color = getSpeakerColor(name);
+      speakerTag = '<div class="rtfc-speaker-tag" style="background:' + color + '">' + escapeHtml(name) + '</div>';
    }
  }

@@ -521,7 +538,13 @@ function buildCard(result) {
  return card;
 }

-function findPendingCard(claim) {
+function findPendingCard(claim, fastClaim) {
+  // try exact match on fast claim first — grounded claim text may differ from fast pass
+  if (fastClaim) {
+    const fastKey = fastClaim.toLowerCase().slice(0, 40);
+    if (pendingCards.has(fastKey)) return pendingCards.get(fastKey);
+  }
+
  const key = claim.toLowerCase().slice(0, 40);
  if (pendingCards.has(key)) return pendingCards.get(key);

@@ -566,6 +589,9 @@ function addVerdict(result) {
  verdictListEl.querySelector('.rtfc-empty')?.remove();
  applyVerdictToBullet(result.claim, result.verdict, result.confidence);
  if (!result._timestamp) result._timestamp = getClaimTimestamp(result.claim);
+  if (result.dominantSpeakerId !== null && result.dominantSpeakerId !== undefined) {
+    result.dominantSpeakerId = String(result.dominantSpeakerId);
+  }
  const card = buildCard(result);
  if (result.pending) {
    const key = result.claim.toLowerCase().slice(0, 40);
@@ -578,11 +604,17 @@ function addVerdict(result) {
 }

 function updateVerdict(result) {
-  const existing = findPendingCard(result.claim);
-  if (!result._timestamp) result._timestamp = getClaimTimestamp(result.claim);
-  // inherit dominantSpeakerId from pending card if grounded result doesn't have one
-  if (existing && existing.dataset.speakerid && !result.dominantSpeakerId) {
-    result.dominantSpeakerId = existing.dataset.speakerid;
+  const existing = findPendingCard(result.claim, result._fastClaim);
+  // always inherit timestamp from the pending card — it was set at detection time
+  // never re-derive from sentenceTimestamps which may no longer contain the original sentence
+  if (existing && existing._resultData?._timestamp) {
+    result._timestamp = existing._resultData._timestamp;
+  } else if (!result._timestamp) {
+    result._timestamp = getClaimTimestamp(result.claim);
+  }
+  // normalize dominantSpeakerId to string for consistent confirmedSpeakerMap lookup
+  if (result.dominantSpeakerId !== null && result.dominantSpeakerId !== undefined) {
+    result.dominantSpeakerId = String(result.dominantSpeakerId);
  }
  const newCard = buildCard(result);
  if (existing) {
@@ -1,7 +1,6 @@
 // session-export.js
 // Handles session logging and PDF export.
-// Loaded after overlay.js — exposes logVerdict(), startSession(), stopSession(), exportPDF() as globals.
-// Reuses escapeHtml() defined in overlay.js (loaded first in manifest.json).
+// Loaded after overlay.js 

 const sessionLog = [];
 let sessionStartTime = null;
@@ -12,7 +12,7 @@ let active = false;

 chrome.runtime.onMessage.addListener((msg, _sender, sendResponse) => {
  if (msg.type === 'START_CAPTURE') {
-    startCapture(msg.streamId)
+    startCapture(msg.streamId, msg.language || 'en')
      .then(() => sendResponse({ ok: true }))
      .catch(err => {
        console.error('[offscreen] error:', err);
@@ -28,8 +28,9 @@ chrome.runtime.onMessage.addListener((msg, _sender, sendResponse) => {
 });

 let utteranceBuffer = '';
+let utteranceSpeakerCounts = {}; // track speaker word counts across buffer chunks

-async function startCapture(streamId) {
+async function startCapture(streamId, language = 'en') {
  if (active) stopCapture();
  active = true;

@@ -51,7 +52,7 @@ async function startCapture(streamId) {
      'sample_rate=16000',
      'channels=1',
      'model=nova-2',
-      'language=en-US',
+      'language=' + language,
      'punctuate=true',
      'interim_results=true',
      'utterance_end_ms=2500',
@@ -83,14 +84,31 @@ async function startCapture(streamId) {
      const text    = result.transcript.trim();
      const isFinal = data.is_final;
      const speech  = data.speech_final;
-      const speaker = result.words?.[0]?.speaker ?? null;
+
+      // accumulate speaker word counts from every chunk
+      if (result.words?.length) {
+        result.words.forEach(w => {
+          if (w.speaker !== null && w.speaker !== undefined) {
+            utteranceSpeakerCounts[w.speaker] = (utteranceSpeakerCounts[w.speaker] || 0) + 1;
+          }
+        });
+      }
+
+      // dominant speaker = whoever had the most words in this utterance so far
+      function getDominantSpeaker() {
+        const entries = Object.entries(utteranceSpeakerCounts);
+        if (!entries.length) return null;
+        return parseInt(entries.sort((a, b) => b[1] - a[1])[0][0]);
+      }

      if (!text) return;

      if (isFinal && speech) {
        // speech_final = end of utterance — send full accumulated text as final
        const fullText = utteranceBuffer ? utteranceBuffer + ' ' + text : text;
+        const speaker  = getDominantSpeaker();
        utteranceBuffer = '';
+        utteranceSpeakerCounts = {};
        chrome.runtime.sendMessage({
          type:    'TRANSCRIPT_RESULT',
          text:    fullText.trim(),
@@ -106,7 +124,7 @@ async function startCapture(streamId) {
          text:    utteranceBuffer,
          isFinal: false,
          interim: true,
-          speaker,
+          speaker: getDominantSpeaker(),
        });
      } else {
        // regular interim — show as-is
@@ -115,7 +133,7 @@ async function startCapture(streamId) {
          text,
          isFinal: false,
          interim: true,
-          speaker,
+          speaker: getDominantSpeaker(),
        });
      }

@@ -175,6 +193,7 @@ function startAudioPipeline() {
 function stopCapture() {
  active = false;
  utteranceBuffer = '';
+  utteranceSpeakerCounts = {};

  if (socket) {
    socket.close();
@@ -81,6 +81,38 @@ body {
  color: #888;
 }

+.key-select {
+  appearance: none;
+  cursor: pointer;
+  background-image: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='10' height='6' viewBox='0 0 10 6'%3E%3Cpath d='M1 1l4 4 4-4' stroke='%23555' stroke-width='1.5' fill='none' stroke-linecap='round'/%3E%3C/svg%3E");
+  background-repeat: no-repeat;
+  background-position: right 8px center;
+  padding-right: 24px;
+}
+
+.key-select option {
+  background: #1a1a1a;
+  color: #e8e8e8;
+}
+
+.lang-row {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+
+.lang-flag {
+  font-size: 18px;
+  line-height: 1;
+  flex-shrink: 0;
+}
+
+.lang-row .key-select {
+  flex: 1;
+}
+
+
+
 .key-hint {
  font-size: 10px;
  color: #555;
@@ -18,6 +18,30 @@
        <label class="key-label">Anthropic API Key</label>
        <input class="key-input" id="anthropicKey" type="password" placeholder="sk-ant-..." autocomplete="off"/>
      </div>
+      <div class="key-field">
+        <label class="key-label">Language</label>
+        <div class="lang-row">
+          <span id="langFlag" class="lang-flag">🇺🇸</span>
+          <select class="key-input key-select" id="languageSelect">
+            <option value="en">en</option>
+            <option value="es">es</option>
+            <option value="fr">fr</option>
+            <option value="de">de</option>
+            <option value="it">it</option>
+            <option value="pt">pt</option>
+            <option value="nl">nl</option>
+            <option value="hi">hi</option>
+            <option value="ja">ja</option>
+            <option value="zh">zh</option>
+            <option value="ar">ar</option>
+            <option value="ko">ko</option>
+            <option value="ru">ru</option>
+            <option value="pl">pl</option>
+            <option value="sv">sv</option>
+            <option value="tr">tr</option>
+          </select>
+        </div>
+      </div>
      <div class="key-hint" id="keyHint"></div>
    </div>

@@ -5,13 +5,27 @@ const statusEl    = document.getElementById('status');
 const anthropicEl = document.getElementById('anthropicKey');
 const keyHint     = document.getElementById('keyHint');
 const keysSection = document.getElementById('keysSection');
+const languageEl  = document.getElementById('languageSelect');
+const langFlagEl  = document.getElementById('langFlag');
+
+const LANG_FLAGS = {
+  en: '🇺🇸', es: '🇪🇸', fr: '🇫🇷', de: '🇩🇪', it: '🇮🇹',
+  pt: '🇧🇷', nl: '🇳🇱', hi: '🇮🇳', ja: '🇯🇵', zh: '🇨🇳',
+  ar: '🇸🇦', ko: '🇰🇷', ru: '🇷🇺', pl: '🇵🇱', sv: '🇸🇪', tr: '🇹🇷',
+};
+
+function updateFlag() {
+  langFlagEl.textContent = LANG_FLAGS[languageEl.value] || '🌐';
+}

 let isActive = false;

-// ── Load saved key ────────────────────────────────────────────────────────────
+// ── Load saved key and language ───────────────────────────────────────────────

-chrome.storage.local.get(['anthropicKey'], (data) => {
+chrome.storage.local.get(['anthropicKey', 'transcriptLanguage'], (data) => {
  if (data.anthropicKey) { anthropicEl.value = data.anthropicKey; anthropicEl.classList.add('saved'); }
+  if (data.transcriptLanguage) languageEl.value = data.transcriptLanguage;
+  updateFlag();
  updateHint();
 });

@@ -27,6 +41,13 @@ anthropicEl.addEventListener('change', () => {
  updateHint();
 });

+// ── Save language on change ───────────────────────────────────────────────────
+
+languageEl.addEventListener('change', () => {
+  chrome.storage.local.set({ transcriptLanguage: languageEl.value });
+  updateFlag();
+});
+
 function updateHint() {
  if (!anthropicEl.value.trim()) {
    keyHint.textContent = 'Enter your Anthropic API key to start.';
@@ -73,8 +94,8 @@ toggleBtn.addEventListener('click', async () => {
    return;
  }

-  // save key then start
-  await new Promise(r => chrome.storage.local.set({ anthropicKey }, r));
+  // save key and language then start
+  await new Promise(r => chrome.storage.local.set({ anthropicKey, transcriptLanguage: languageEl.value }, r));

  chrome.runtime.sendMessage({ type: 'START_FACTCHECK' }, (res) => {
    if (res?.ok) {