Auto-Question-Generation-on-Dynamic-Programming-LCS-1.pptx
- 1. Auto Question -Answer Generation on Dynamic Programming MAHATMA GANDHI MISSION’S COLLEGE OF ENGINEERING, NANDED GUIDED BY: DR. KAPRE MAM PRESENTED BY : TEJAS KOTALWAR (218) SOHAM KASHETTIWAR (225) GANESH KONDAMWAR (224) AKSHRAPLUS (INTERNSHIP)
- 2. Introduction to Dynamic Programming What is Dynamic Programming (DP)? Dynamic Programming is a technique to solve complex problems by breaking them down into simpler overlapping subproblems and storing their results to avoid redundant computations. Applications It's widely applied in various fields. • Sequence analysis • Computational biology • Text processing Key Features of DP: Overlapping Subproblems: Problems can be broken down into smaller, repeated subproblems. Optimal Substructure: An optimal solution can be constructed from optimal solutions of subproblems. Memoization or Tabulation: Used to store subproblem results (Top-down or Bottom-up approach).
- 3. Subtopics of Dynamic Programming Overlapping subproblems and optimal substructure Longest Common Subsequence (LCS) Fibonacci sequence and memoization Matrix Chain Multiplication 0/1 Knapsack Problem Floyd-Warshall Algorithm
- 4. What is LCS (Longest Common Subsequence)? LCS is the longest sequence that appears in the same relative order (but not necessarily contiguous) in both strings. Example: •String X: ABCBDAB •String Y: BDCAB •LCS: BCAB or BDAB (Length = 4) LCS Characteristics: •Not a substring (can skip characters). •Has optimal substructure → perfect for dynamic programming. •Frequently used in: • DNA sequence alignment • File diff tools • Plagiarism detection • Code similarity checkers Time Complexity: •O(m * n) using DP table (where m and n are lengths of the two strings)
- 5. Project Objective Why Auto Question Generation? Why Auto Question Generation? •To reduce manual effort in creating diverse and meaningful coding questions. •To support automated learning tools, mock tests, and online coding platforms. •Useful for students, educators, and interview preparation platforms. Our Goal: Build an intelligent system that can automatically generate unique, high-quality questions related to LCS across different categories (conceptual, implementation, algorithmic) and difficulty levels (easy, medium, hard).
- 6. Dataset Creation Custom Dataset Built for LCS Question Generation Custom Dataset Built for LCS Question Generation •Designed a template-based dataset using JSON. •Each entry includes: •Template: Text format with placeholders {X}, {Y} •Category: Conceptual | Implementation | Algorithmic •Difficulty: Easy | Medium | Hard •Variables: List of possible values for placeholders
- 7. T5 Model Overview 1 Transformer-Based A powerful model designed for text-to-text tasks. 2 Fine-Tuned Specifically for generating questions on coding concepts related to LCS. 3 Dataset Curated from platforms like HackerRank and LeetCode.
- 8. System Workflow Input Provide keywords and a difficulty level for the question. Processing The system preprocesses data and fine-tunes the T5 model. Output The system generates and refines questions with test cases.
- 10. Sample Questions Generated ✅ Q1: "Write a recursive function to compute the LCS of AB and AX." ✅ Q2: "Find the length of the longest common subsequence between the strings HELLO and WORLD." ✅ Q3: "What is the LCS of MNOP and PROGRAM?" ✅ Q4: "Determine the longest common subsequence of DYNAMIC and LANGUAGE."
- 11. Answer Generation Answer Generation using BERT Model Objective: Automatically understand the type of LCS question and generate the correct answer or response using classification + logic. How It Works: 1. 🧾 Input: LCS-related question e.g., “Find the LCS of apple and maple.” 2. 🧠 Question Classification using BERT •Fine-tuned a BERT-based classifier to detect the type of question: •Basic LCS •Only Length •Multiple Pairs •Incomplete/Vague Question 3. 🔁 Function Mapping •Each category maps to a function from lcs_solver.py: •solve_basic_lcs() – Extracts and computes full LCS •solve_lcs_length_only() – Returns only LCS length •solve_multiple_pairs() – Handles multi-comparison •solve_incomplete_question() – Gracefully handles vague inputs
- 12. Answer Generation Answer Generation using BERT Model 4. ✅ Answer Output Example: Q: "What is the LCS of apple and maple?" A: 💡 The Longest Common Subsequence is ‘aple’, with length 4.
- 13. Conclusion and Future Scope AI Automation AI automates tasks for scalable educational tools. Future Extension Extend to other dynamic programming problems. Multilingual Support Incorporate multilingual support and interactive components.