Hello everyone, I'm working on a project to develop an autonomous AI system that predicts upcoming exam papers based on previous years' papers. The system will incorporate pattern recognition to analyse the types of questions, considering the exam syllabus, the number of questions, and the different formats. This approach aims to predict new questions that may appear in future exams.
Rough Approach :
Optical Character Recognition(OCR) to convert scanned exam papers into text format.
Question Pattern Analysis: NLP techniques are used to analyse the pattern in questions.
Question Generation: New question are generated using Machine Learning(ML) and Deep Learning(DL) models.
Model Evaluation : Lastly model is evaluated on accuracy.
Discussion and Request for Suggestions:
One aspect I'm considering is the use of a pretrained AI model to enhance the performance of the system, given that I have a relatively small dataset. I believe that Google's Flan-T5 could be a strong candidate for this project.
I've outlined my approach and ideas, and I'm eager to receive suggestions or feedback. I hope to embark on this project collaboratively with others
who might be interested.
Email: larrylab2001@tutanota.com
Top comments (2)
Your project sounds fascinating and has great potential for helping students prepare more effectively. Using OCR, NLP for pattern analysis, and ML for question generation is a smart approach. Leveraging a pretrained model like Google's Flan-T5 could definitely enhance the system, especially with a smaller dataset. It would be interesting to discover more ways in which collaboration can refine the model's accuracy and performance further. Best of luck with the development!
🤣😂