Quizify: Enriching Learning Through Open-Source Models
The Spark
Meet Quizify, a groundbreaking interactive learning tool set to redefine studying as you know it. Our vision is grounded in our shared passion for building distinctive, cutting-edge applications that take the monotony out of learning.
The Functionality
Quizify is an advanced application that puts the power of creating compelling quizzes in the hands of the user. By harnessing the capabilities of open-source models from Hugging Face, Quizify breathes life into interactive multiple-choice quizzes based on user input or content harvested from a given Wikipedia URL.
The Building Blocks
Our team - a vibrant mix of data scientists, engineers, and analysts - joined hands to construct Quizify. We deployed Python, Flask, JavaScript, CSS, and HTML to build an intuitive frontend interface, ensuring a fluid and user-friendly navigation experience. The open-source models from Hugging Face served as our foundation, facilitating the creation of captivating questions and distractors, which enhance the user learning experience. Databricks notebooks became our research companions, helping us refine our algorithms for improved accuracy and efficiency.
The Language Learning Model (LLM), an open-source model, proved integral to our process. We used the Question Answer Generation model (potsawee/t5-large-generation-squad-QuestionAnswer) and the Distractor Generation model (potsawee/t5-large-generation-race-Distractor) from LLM. These models enable us to generate stimulating questions and distractors that enrich the learning journey for users.
The Hurdles
Building Quizify wasn't without its challenges. Crafting a dynamic, responsive, and user-friendly interface was a tough nut to crack. But with careful planning and dedicated UI/UX design, we shaped an experience that is intuitive and seamless for our users.
Another significant obstacle was finding the right open-source model that served our purpose. Many models needed heavy fine-tuning or did not perform up to our expectations. We faced errors, and some models were simply too bulky to run given our computational constraints.
However, Databricks came to our rescue. With its powerful GPUs and computing capabilities, it allowed us to sift through various models, test them, and ultimately find the ones that ticked all our boxes, overcoming the initial setbacks.
Triumphs We Relish
We're thrilled about our accomplishments, one of which includes enhancing the app's performance through rigorous data cleaning, text extraction, and context analysis based on the open-source model. This has made the app more efficient and accurate, a testament to our team's hard work and dedication.
We're equally proud of Quizify's user-friendliness. The app is designed to simplify quiz creation, making it accessible for users with varying technical skills. We believe learning should be for everyone, and Quizify reflects this principle.
Additionally, this journey has been a series of firsts for us: building an interactive app, venturing into video editing, designing user interfaces, and even employing the LLM locally. Each 'first' brought its own set of challenges, but we rose to the occasion and emerged stronger.
Lessons Gathered
The Quizify journey was a treasure trove of learning. Technically, we forayed into new areas: interactive app development, video editing, user interface design, and running the LLM locally. Each new exploration boosted our technical acumen.
But the learning wasn't limited to tech. The project taught us invaluable lessons in team dynamics. Effective communication emerged as a critical factor, helping us align our goals, resolve issues, and work efficiently. Moreover, we discovered the impact of motivation, the force that kept us going in the face of challenges. In essence, this journey helped us refine our teamwork
Quizify's Future Trajectory
Our ambitions for Quizify stretch beyond this competition. We envision it as a universally useful tool.
Future enhancements include diverse question types like True/False and Matching, and a feature for text extraction from images. We plan to increase its capacity to handle larger texts like textbooks via an import PDF feature and introduce varied quiz difficulty levels. We're also developing a feature to allow users to divide text into sections for focused quizzes.
Our journey with Quizify is ongoing, driven by our commitment to continual innovation and learning. We aim for Quizify to become an increasingly engaging and effective learning tool. The path of learning never ends, and neither does our pursuit of creating the best learning experiences.
Log in or sign up for Devpost to join the conversation.