A stressful and taxing continuous evaluation methodology presented itself during this global pandemic which inspired us to create a solution to this very problem.
What it does
Project AQUA is a platform that will organize the system of assignments and quizzes conducted by schools/colleges for the continuous evaluation of students.
👩🏫 We provide the following services: 🤖 AI powered Question Generator 🤓 Keyword Checker 🤓 Plagiarism Checker 🔖 Notice Board ☁️ Assignment Upload/Download Portal 🙃 Anti Burden System
How we built it
- We used SQuADv1 dataset and and T5 model with highlight preprocessing format. This idea is proposed in the "A Recurrent BERT-based Model for Question Generation"
- We used TF-IDF to vectorize the documents and then calculated the cosine similarity between them to check for potential plagiarism
- We used React and Context (for managing global state) on client side to make a single page application with a good UI.
- We used Express, MongoDB on server side to store the details of users, notices and assignments with their url returned by the assignment files which were being stored on Azure's blob storage. -
Challenges we ran into
- Fixing CORS was a quite okay challenge that we ran into as we unexpectedly got stuck and couldn't fix it easily.
- Setting up a cloud storage pipeline such that it was available to both MongoDB and Azure's Blob storage.
Accomplishments that we're proud of
- Making an efficient and robust database schema ensuring that database queries are optimized.
- Building a cloud storage pipeline for file handling accessible to both MongoDB and Azure's Blob storage.
- Training an NLP model from scratch.
What we learned
- Azure Cognitive Services which allowed us to easily integrate AI-based features into our application.
- How to keep all the data sources consistent.
What's next for AQUA
We want our solution to replacing existing makeshift solutions to this problem so that it can be used by the students and teachers actively in the institutions.