Banca Intelligence (AI Relevance Analysis)

Overview

Banca Intelligence uses Natural Language Processing (NLP) and fuzzy matching algorithms to analyze which topics from your exam syllabus are most likely to appear on the actual exam, based on historical incidence data from the exam board (banca).

This feature uses advanced text similarity algorithms including Levenshtein distance, tokenization, and stopword filtering to match your syllabus topics against known high-incidence themes.

How It Works

1. Topic Matching Pipeline

Your Syllabus Topic
      ↓
  Tokenization & Normalization
      ↓
  Fuzzy Match Against Hot Topics Database
      ↓
  Calculate Relevance Score (0-100)
      ↓
  Assign Priority (P1, P2, P3)

2. NLP Engine

The system uses multiple techniques to find matches:

Tokenization

Breaks text into meaningful words, removes stopwords:

// Tokenization (relevance.js:21-30)
export function tokenize(text) {
  if (!text) return [];
  
  const normalized = text.toLowerCase()
    .normalize('NFD').replace(/[\u0300-\u036f]/g, '') // Remove accents
    .replace(/[^\w\s]/gi, ' ') // Remove punctuation
    .replace(/\s+/g, ' ')
    .trim();

  return normalized.split(' ').filter(word => 
    word.length > 2 && !STOPWORDS.has(word)
  );
}

Stopwords (ignored words):

const STOPWORDS = new Set([
  'o', 'a', 'os', 'as', 'de', 'do', 'da', 'em', 'no', 'na',
  'para', 'com', 'sem', 'e', 'ou', 'que', 'se',
  'lei', 'artigo', 'art', 'inciso', 'capítulo',
  'conceito', 'noções', 'introdução', 'teoria', 'geral'
]);

Fuzzy Similarity (Levenshtein Distance)

Calculates how similar two words are:

// Levenshtein distance (relevance.js:33-53)
export function levenshteinDistance(a, b) {
  if (a.length === 0) return b.length;
  if (b.length === 0) return a.length;

  const matrix = Array(a.length + 1).fill(null).map(() => 
    Array(b.length + 1).fill(null)
  );

  for (let i = 0; i <= a.length; i++) matrix[i][0] = i;
  for (let j = 0; j <= b.length; j++) matrix[0][j] = j;

  for (let i = 1; i <= a.length; i++) {
    for (let j = 1; j <= b.length; j++) {
      const cost = a[i - 1] === b[j - 1] ? 0 : 1;
      matrix[i][j] = Math.min(
        matrix[i - 1][j] + 1,      // deletion
        matrix[i][j - 1] + 1,      // insertion
        matrix[i - 1][j - 1] + cost // substitution
      );
    }
  }
  
  return matrix[a.length][b.length];
}

// Similarity percentage (relevance.js:56-61)
export function fuzzySimiliarity(word1, word2) {
  const dist = levenshteinDistance(word1, word2);
  const maxLen = Math.max(word1.length, word2.length);
  if (maxLen === 0) return 1.0;
  return (maxLen - dist) / maxLen; // 0.0 to 1.0
}

Token Matching

Compares entire phrases:

// Token set matching (relevance.js:64-81)
export function computeTokenMatch(tokensA, tokensB) {
  if (tokensA.length === 0 || tokensB.length === 0) return 0;

  let matches = 0;
  
  for (const ta of tokensA) {
    let bestWordScore = 0;
    for (const tb of tokensB) {
      const sc = fuzzySimiliarity(ta, tb);
      if (sc > bestWordScore) bestWordScore = sc;
    }
    
    if (bestWordScore > 0.8) { // 80% similarity threshold
      matches++;
    }
  }
  
  return matches / Math.max(tokensA.length, tokensB.length);
}

3. Best Match Algorithm

Finds the most relevant hot topic for each syllabus item:

// Best match finder (relevance.js:92-163)
export function findBestMatch(editalSubjectName, disciplinaId = null) {
  const hotTopics = state.bancaRelevance?.hotTopics || [];
  const userMap = state.bancaRelevance?.userMappings?.[editalSubjectName];

  // 1. Manual override check
  if (userMap) {
    if (userMap === 'NONE') {
      return { 
        matchedItem: null, 
        score: 0.05, 
        confidence: 'HIGH', 
        reason: 'Marcado como sem incidência pelo usuário' 
      };
    }
    const forcedMatch = hotTopics.find(h => h.id === userMap);
    if (forcedMatch) {
      return { 
        matchedItem: forcedMatch, 
        score: 1.0, 
        confidence: 'HIGH', 
        reason: 'Mapeamento Fixado Manualmente' 
      };
    }
  }

  // 2. Automatic matching
  const tokensA = tokenize(editalSubjectName);
  const strA = editalSubjectName.toLowerCase().trim();

  let bestMatch = null;
  let highestScore = 0;
  let bestReason = '';
  let bestConf = 'LOW';

  for (const ht of hotTopics) {
    const strB = ht.nome.toLowerCase().trim();

    // Exact match
    if (strA === strB) {
      return { 
        matchedItem: ht, 
        score: 1.0, 
        confidence: 'HIGH', 
        reason: 'Match Exato de Título' 
      };
    }

    // Substring match
    if (strA.includes(strB) || strB.includes(strA)) {
      const sc = 0.9;
      if (sc > highestScore) {
        highestScore = sc;
        bestMatch = ht;
        bestReason = 'Contém Nome Parcial';
        bestConf = 'HIGH';
      }
    }

    // Token-based fuzzy match
    const tokensB = tokenize(ht.nome);
    const tokenScore = computeTokenMatch(tokensA, tokensB);

    if (tokenScore > highestScore) {
      highestScore = tokenScore;
      bestMatch = ht;
      
      if (tokenScore > 0.8) {
        bestConf = 'MEDIUM';
        bestReason = 'Termos Altamente Similares (Algoritmo)';
      } else {
        bestConf = 'LOW';
        bestReason = `Plausível Semelhança Textual (${Math.round(tokenScore * 100)}%)`;
      }
    }
  }

  // No good match found
  if (highestScore < 0.4) {
    return { 
      matchedItem: null, 
      score: 0.05, 
      confidence: 'HIGH', 
      reason: 'Sem incidência detectada nas chaves da Banca' 
    };
  }

  return { matchedItem: bestMatch, score: highestScore, confidence: bestConf, reason: bestReason };
}

4. Relevance Score Calculation

// Final score calculation (relevance.js:169-214)
export function calculateFinalRelevance(editalSubjectCtx) {
  const matchStruct = findBestMatch(
    editalSubjectCtx.assuntoNome, 
    editalSubjectCtx.disciplinaId
  );
  
  if (!matchStruct.matchedItem) {
    return { finalScore: 0.0, priority: 'P3', matchData: matchStruct };
  }

  const ht = matchStruct.matchedItem;

  // A. Incidence score (how often the topic appears in exams)
  let incidenceScore = 0;
  
  if (ht.weight !== undefined) {
    incidenceScore = ht.weight; // 0-1 scale
    if (incidenceScore > 1) incidenceScore = incidenceScore / 100;
  } else if (ht.rank !== undefined) {
    // Rank #1 = highest weight, decays progressively
    incidenceScore = 1 / Math.pow(ht.rank, 0.7);
  } else if (ht.level !== undefined) {
    // 'ALTA' = 1.0, 'MEDIA' = 0.6, 'BAIXA' = 0.3
    incidenceScore = ht.level === 'ALTA' ? 1.0 : (ht.level === 'MEDIA' ? 0.6 : 0.3);
  }

  // B. Combine incidence with match quality
  let finalScore = incidenceScore * matchStruct.score;

  // C. Bonus for weak areas (more errors = higher priority)
  if (editalSubjectCtx.erros / (editalSubjectCtx.acertos + editalSubjectCtx.erros + 1) > 0.6) {
    finalScore += 0.1;
  }

  // D. Assign priority tier
  let priority = 'P3';
  if (finalScore >= 0.75) priority = 'P1';
  else if (finalScore >= 0.40) priority = 'P2';

  return {
    finalScore: Math.round(finalScore * 100), // 0-100 scale
    priority,
    matchData: matchStruct
  };
}

Priority Tiers

Priority	Score Range	Color	Meaning
P1	75-100	Red	Critical - High incidence, study immediately
P2	40-74	Yellow	Important - Moderate incidence, study soon
P3	0-39	Gray	Low - Rare or no incidence, deprioritize

Applying Ranking to Edital

// Apply ranking to entire exam (relevance.js:217-257)
export function applyRankingToEdital(editalId) {
  const edt = state.editais.find(e => e.id === editalId);
  if (!edt) return [];

  let flatList = [];

  // Calculate score for each topic
  edt.disciplinas.forEach(d => {
    d.assuntos.forEach(a => {
      const result = calculateFinalRelevance({
        assuntoNome: a.nome,
        disciplinaId: d.id,
        acertos: a.acertos || 0,
        erros: a.erros || 0
      });

      flatList.push({
        discId: d.id,
        discNome: d.nome,
        assuntoId: a.id,
        assuntoNome: a.nome,
        ...result
      });
    });
  });

  // Sort by score (highest first)
  flatList.sort((a, b) => b.finalScore - a.finalScore);

  // Reassign priorities based on percentile
  const top20Index = Math.max(0, Math.floor(flatList.length * 0.2) - 1);
  const top60Index = Math.max(0, Math.floor(flatList.length * 0.6) - 1);

  flatList.forEach((item, index) => {
    if (index <= top20Index) item.priority = 'P1';
    else if (index <= top60Index) item.priority = 'P2';
    else item.priority = 'P3';
  });

  return flatList;
}

Committing the Ranking

Saves the calculated priorities to the database:

// Persist ranking (relevance.js:260-304)
export function commitEditalOrdering(editalId, rankedFlatList) {
  const edt = state.editais.find(e => e.id === editalId);
  if (!edt) return false;

  // Group by discipline
  const grouped = {};
  rankedFlatList.forEach(item => {
    if (!grouped[item.discId]) grouped[item.discId] = [];
    grouped[item.discId].push(item);
  });

  // Reorder topics within each discipline
  edt.disciplinas.forEach(d => {
    const sortedItemsForDisc = grouped[d.id] || [];
    const newAssuntosArray = [];
    
    sortedItemsForDisc.forEach(sItem => {
      const originalAssunto = d.assuntos.find(a => a.id === sItem.assuntoId);
      if (originalAssunto) {
        // Inject relevance metadata
        originalAssunto.relevance = {
          priority: sItem.priority,
          finalScore: sItem.finalScore,
          reason: sItem.matchData.reason,
          confidence: sItem.matchData.confidence
        };
        newAssuntosArray.push(originalAssunto);
      }
    });

    // Preserve any topics that weren't ranked
    d.assuntos.forEach(a => {
      if (!newAssuntosArray.find(na => na.id === a.id)) {
        a.relevance = { 
          priority: 'P3', 
          finalScore: 0, 
          reason: 'Não Rankeado', 
          confidence: 'LOW' 
        };
        newAssuntosArray.push(a);
      }
    });

    d.assuntos = newAssuntosArray;
  });

  scheduleSave();
  return true;
}

Using Banca Intelligence

Set Up Hot Topics Database

Navigate to Inteligência de Banca and click “Gerenciar Base de Dados”. Add known high-incidence topics for your exam board.

state.bancaRelevance = {
  hotTopics: [
    { 
      id: 'ht_1', 
      nome: 'Direitos Fundamentais', 
      weight: 0.95, 
      disciplinaId: null 
    },
    { 
      id: 'ht_2', 
      nome: 'Controle de Constitucionalidade', 
      rank: 2, 
      disciplinaId: null 
    }
  ],
  userMappings: {}
};

Run Analysis

Select an edital and click “Analisar Relevância”. The system matches each syllabus topic against the hot topics database.

Review Results

The analysis shows:

Match quality (Exact, High, Medium, Low)
Relevance score (0-100)
Priority tier (P1, P2, P3)
Matching reason

Manual Adjustments (Optional)

Override incorrect matches by clicking “Corrigir” and selecting the correct hot topic or marking as “Sem Incidência”.

// Manual mapping (stored in state)
state.bancaRelevance.userMappings['Direitos Sociais'] = 'ht_5';
// or
state.bancaRelevance.userMappings['Tópico Irrelevante'] = 'NONE';

Commit Ranking

Click “Salvar Ordenação” to apply the ranking permanently. Topics are reordered by priority within each discipline.

Study by Priority

Focus on P1 topics first. The Study Organizer and Calendar can filter by priority to optimize your study plan.

Banca Intelligence is only as good as your hot topics database. Ensure you have accurate historical incidence data for best results.

Reverting Rankings

Restore alphabetical order and remove priority tags:

// Revert to alphabetical order (relevance.js:307-324)
export function revertEditalOrdering(editalId, disciplinaId) {
  const edt = state.editais.find(e => e.id === editalId);
  if (!edt) return false;

  const disc = edt.disciplinas.find(d => d.id === disciplinaId);
  if (!disc) return false;

  // Remove relevance metadata
  disc.assuntos.forEach(a => {
    delete a.relevance;
  });

  // Sort alphabetically
  disc.assuntos.sort((a, b) => a.nome.localeCompare(b.nome));

  scheduleSave();
  return true;
}

Best Practices

Curate Hot Topics Carefully

Only add topics with verified high incidence. Garbage in = garbage out. Use official exam board statistics if available.

Use Exact Topic Names

Match wording from official exam notices as closely as possible. This improves fuzzy matching accuracy.

Review Low Confidence Matches

Always check matches with “LOW” confidence. These often need manual correction.

Combine with Performance Data

The algorithm boosts priority for topics where you have high error rates. This ensures weak areas get attention.

Update Regularly

Re-run analysis after adding new hot topics or when exam boards release updated statistics.

Get Started

Core Features

Study Planning

Synchronization

Configuration

Banca Intelligence (AI Relevance Analysis)

Overview

How It Works

1. Topic Matching Pipeline

2. NLP Engine

Tokenization

Fuzzy Similarity (Levenshtein Distance)

Token Matching

3. Best Match Algorithm

4. Relevance Score Calculation

Priority Tiers

Applying Ranking to Edital

Committing the Ranking

Using Banca Intelligence

Reverting Rankings

Best Practices

Build docs developers (and LLMs) love

Get Started

Core Features

Study Planning

Synchronization

Configuration

​Overview

​How It Works

​1. Topic Matching Pipeline

​2. NLP Engine

​Tokenization

​Fuzzy Similarity (Levenshtein Distance)

​Token Matching

​3. Best Match Algorithm

​4. Relevance Score Calculation

​Priority Tiers

​Applying Ranking to Edital

​Committing the Ranking

​Using Banca Intelligence

​Reverting Rankings

​Best Practices

Build docs developers (and LLMs) love

Overview

How It Works

1. Topic Matching Pipeline

2. NLP Engine

Tokenization

Fuzzy Similarity (Levenshtein Distance)

Token Matching

3. Best Match Algorithm

4. Relevance Score Calculation

Priority Tiers

Applying Ranking to Edital

Committing the Ranking

Using Banca Intelligence

Reverting Rankings

Best Practices