SmartEval
SmartEval: Empowering Academic Excellence with AI-Powered Assessment
Created on 16th July 2023
•
SmartEval
SmartEval: Empowering Academic Excellence with AI-Powered Assessment
The problem SmartEval solves
Academic life involves various tasks such as creating and evaluating educational content, assessing student performance, providing feedback, and managing administrative tasks. This is a significant challenge faced by both students and educators. However, traditional assessment methods often pose several challenges that can hinder the efficiency and effectiveness of the academic process.
Traditional assessment methods are quite time consuming and labor-intensive and have a limited scalability.
Here comes the need for automated tools in the education industry to streamline the evaluation process. These tools leverage machine learning and artificial intelligence techniques to provide efficient and objective assessments, benefiting both students and educators.
Challenges we ran into
There were multiple challenges we ran into, here are some of them which are worth stating:
- The utilization of a single keyword extraction technique proved insufficient, as it often produced redundant keywords. To address this, we skillfully integrated two architectures, leveraging their combined outputs for more accurate and realistic keyword extraction. This approach ensured the extraction of the most relevant and meaningful keywords, enhancing the precision of our evaluation process.
- Creating a modular code structure was a significant challenge during the development process. However, we found a solution by embracing the principles of Object-Oriented Programming (OOPs). This modular approach improves code readability, maintainability, and scalability, making it easier to enhance and expand our tool's capabilities in the future.
- Resolving library errors were, though not so much, but time-consuming.
- Resources were always an issue, as we all know, ML models are bulky, and take a hell lot of computational power as well as GPU RAM to execute, and having limited resources, we did not had the privilege of implementing our models on local systems, and had to use colab for most of the tasks.
- Interfacing was yet again a crucial step, but having tried integrating python code using traditional web-builders like HTML, CSS, NODEJS, etc, we had to use tkinter, yet again library implementation of python.
But, then again, we were not able to host our built interface, but then, found a way out by converting the code into executable files which can directly be run into any system.