Exam Question Pattern Analyzer

The Exam Question Pattern Analyzer uses pure client-side algorithms to analyze past exam papers and predict high-probability questions. It extracts questions from text, PDF, or images, detects topics using keyword matching, calculates question frequency using Jaccard similarity, assigns difficulty levels and Bloom's taxonomy classifications, and generates year-wise trend analysis—all without any external API calls.

Our prediction engine uses a weighted scoring formula: Importance Score = (Repeat Count × 0.4) + (Recent Year Weight × 0.3) + (High Marks × 0.2) + (Hard Difficulty × 0.1). Questions that appear frequently, in recent exams, carry high marks, and are classified as difficult receive the highest prediction scores. The top 10 are displayed as your most likely exam questions.

We provide built-in keyword dictionaries for: Operating Systems (processes, memory, deadlock, synchronization), DBMS (SQL, normalization, transactions), OOP (inheritance, polymorphism, design patterns), Data Structures (arrays, trees, graphs, sorting), Computer Networks (OSI, TCP/IP, routing), and Software Engineering (SDLC, testing, requirements). You can also add custom topics with your own keywords for any subject.

Difficulty is determined by analyzing action verbs in questions: Easy questions use words like "define", "list", "state", "name". Medium questions include "explain", "describe", "compare", "discuss". Hard questions contain "analyze", "design", "evaluate", "derive". The marks assigned to a question also factor into the difficulty score—10+ mark questions lean toward "hard".

Bloom's Taxonomy classifies learning into six cognitive levels: Remember (recall facts), Understand (explain concepts), Apply (use knowledge), Analyze (break down information), Evaluate (make judgments), and Create (produce new work). Our tool maps each question to these levels based on action verb keywords, helping you understand what type of thinking your exam tests most.

Yes! The tool uses PDF.js to extract text from PDF files directly in your browser. Simply drag and drop your PDF files into the upload zone, and the text will be extracted automatically. For scanned PDFs (images), you can use the OCR tab with Tesseract.js for optical character recognition—all processed locally without uploading to any server.

We use Jaccard Similarity to compare questions word-by-word. Questions that share more than your set threshold (default 70%) of words are grouped as "similar". This catches not just identical questions, but rephrased versions asking the same concept. You can adjust the similarity threshold to be more or less strict based on your needs.

Based on your analysis results, the Study Plan Generator creates a personalized day-by-day study schedule. It ranks topics by importance (combining frequency and marks weight), then allocates study hours proportionally. High-priority topics get more time. Simply enter your days until exam and hours per day, and the tool generates a structured plan with time allocated for each topic.

100% Private. All analysis—text extraction, topic detection, frequency matching, prediction scoring—happens locally in your browser using JavaScript. Your question papers are never uploaded to any server. PDF parsing, OCR, and all algorithms run client-side. This makes it completely safe for analyzing confidential exam materials or institutional papers.

Yes! You can export your complete analysis as: PDF Report (printable document with all charts and predictions), CSV File (spreadsheet with all questions, marks, topics, and difficulty), or Study Sheet (quick reference text file with important topics and predicted questions). All exports are generated locally without any server processing.