about project



In the realm of modern education, assessing the cognitive demand of questions is crucial for designing effective learning experiences. The primary goal of the thesis is to explore three hypothesis (a) Whether transformer-based text representation schemes combined with different classification methods can provide superior performance in question classification according to Bloom's Taxonom ? (b) Can short text question can be clustered as per Bloom's Taxonomy level by using transformer based text representation technique combined with different clustering algorithms ? (c) Can the expansion of knowledge through clustering can provide better results in classification ? So that the power of machine learning and natural language processing, can be used to automatically identify the cognitive complexity of questions based on established educational frameworks. This initiative aims to bridge the gap between pedagogical theory and AI-driven automation in education.

highlights



Automatic Deep Question Categorization using Classification


This work presents an automated approach to categorizing educational questions using deep learning-based classification models. By leveraging advanced language models, the system analyzes and assigns cognitive levels to questions, enabling scalable and consistent evaluation of learning materials. The method bridges artificial intelligence and pedagogy, offering a reliable solution for classifying questions based on their complexity and intent.

Automatic Deep Question Categorization using Clustering


This work explores the use of deep learning-based representations combined with unsupervised clustering techniques to automatically categorize educational questions. By transforming questions into contextual embeddings and applying advanced clustering algorithms, the system uncovers latent cognitive structures without the need for labeled data. This approach offers a scalable solution for organizing and analyzing large sets of educational content based on their underlying complexity.

Automatic Deep Question Categorization with integration of Classification and Clustering


This work presents a hybrid framework that integrates deep learning-based classification with unsupervised clustering to automatically categorize educational questions. By combining the strengths of both approaches, the system improves accuracy in identifying cognitive levels while enabling the labeling of previously unseen or unlabeled questions. The integration enhances scalability, adaptability, and consistency in evaluating educational content based on question complexity.

Automatic Deep Question Categorization a Web Application


This project delivers a user-friendly web application that automates the categorization of educational questions using deep learning techniques. Users can upload question papers or input individual questions to receive instant classification based on cognitive levels. Powered by a backend integration of transformer models and intelligent clustering, the application provides scalable, real-time support for educators seeking to analyze and organize assessment content efficiently.

Publications



CODE



The algorithm is implemented in Python. The code and datasets are available in the following links.

Use of code/technique/algorithm is free as long as it is used for any academic and non-commercial purpose. If you use this code/technique/algorithm, please cite this work.

For any query regarding the algorithms, please mail to indrajit@nitttrkol.ac.in

Disclaimer:
The dataset is used from public database. Thus, NITTTR, Kolkata does not own any responsible for its accuracy.

MEMBERS



Avatar

Dr. Indrajit Saha

Prjoect Supervisor

Avatar

Aritra Mukherjee

M.Tech Student

Avatar

Alakh Yadav

Research Assistant