LearnTerms v2: Smarter Studying, Simplified

28 February 2024
Justin Dang
Justin A. Dang NSU Oklahoma College of Optometry Class of 2028
LearnTerms v2
Smarter Studying, Simplified.

Introduction

Welcome to the launch of LearnTerms v2 – a revolutionary upgrade designed to transform your exam preparation. With smarter study tools and an AI-powered backend, LearnTerms v2 is set to become the go-to platform for our class. In this post, we'll walk you through the evolution of the platform, explain the technology behind the upgrade, and show you how these changes will make your study sessions more efficient and effective.

Background & The Problem

Background

LearnTerms began as a focused solution addressing a specific challenge in a single course. By concentrating our efforts on a niche area, we were able to create a lightweight and user-friendly service with minimal barriers to entry. However, as our user base grew and expectations increased, the limitations of the original version became apparent.

The Problem

  • Limited Scope: The previous version was confined to a single class, which restricted its potential.
  • Manual Verification: A significant amount of manual effort was required to verify terms and maintain accuracy, consuming countless hours.
  • Scalability Challenges: The system needed to evolve to accommodate a broader range of content while still delivering a high-quality user experience.

These challenges highlighted the need for a robust, scalable solution that would automate and streamline study material generation.

The Evolution: What's New in LearnTerms v2

Vision for v2

LearnTerms v2 is built with three clear objectives:

  • Robust Platform: Develop a study tool that is not only powerful but also incredibly easy to use.
  • Rapid and Accurate Content Generation: Utilize AI to produce high-quality, accurate study questions quickly.
  • Personalized Learning: Maintain domain-specific expertise to ensure the content is tailored to your course and study habits.

Feedback from our beta phase was critical in shaping these upgrades, ensuring that each new feature addresses real-world study challenges.

User Benefits

  • Enhanced Efficiency: Experience faster load times and smoother interactions.
  • Reliable Exam Preparation: Access a consistent and dependable exam module that simulates real test conditions.
  • Cross-Device Continuity: Enjoy persistent saving across devices so you can pick up exactly where you left off.

Under the Hood: Upgrading Our Core Technologies

Svelte 5: Speed and Efficiency

Upgrading from Svelte 4 to Svelte 5 has dramatically improved the platform. Key benefits include:

  • Faster Initial Load: Reduced waiting time when you first access the platform.
  • Smoother Interactions: Enhanced responsiveness for a more fluid user experience.
  • Improved Developer Experience: New features streamline the coding process, translating into quicker updates and bug fixes.

LearnTerms Exam Module First Load Performance

Lower is better

Svelte 5 ~ 0.9 s

Svelte 4 ~ 3.4 s

The new Exam Module's first load time sees a 75% improvement on average when comparing Svelte 5 to Svelte 4.

daisyUI v5: A Fresh and Modern Look

The new daisyUI v5 update brings a modern, refined design to LearnTerms. Highlights include:

  • Lean and Lightweight: The library is now 61% smaller and dependency-free.
  • Enhanced Aesthetics: A cleaner, more mature look that still retains the playful essence of the original design.
  • Quick Deployment: Simplifies the process of building and deploying new features, ensuring a consistent visual experience.

TailwindCSS 4.0: Performance and Modern Standards

With the upgrade to TailwindCSS 4.0, we see:

  • Faster Builds: Both full and incremental build times have significantly decreased.
  • Optimized Styling: The high-performance engine ensures our custom styles load quickly and render beautifully.
  • Modern Standards: Continued support for modern web standards enhances overall performance and compatibility.

Benchmark Comparison

v3.4v4.0Improvement
Full build378ms100ms3.78x
Incremental rebuild with new CSS44ms5ms8.8x
Incremental rebuild with no new CSS35ms192µs182x

Lower numbers indicate better performance.

Harnessing the Power of AI for Personalized Learning

Introduction to AI Integration

LearnTerms v2 leverages advanced Large Language Models (LLMs) to generate personalized study questions quickly and accurately. By fine-tuning these models and utilizing effective prompt engineering, we ensure that every generated question is both relevant and accurate.

The AI Workflow

  1. Define the Task:

    Clear instructions guide the model on what is expected. For example, a prompt might direct the model to generate diverse question types that challenge understanding and critical thinking.

  2. Specify the Format:

    We provide an exact format for the questions, including:

    • A numbered question format
    • Multiple-choice options
    • An explanation section that details why certain options are correct or incorrect

    Example Prompt Block:

    "You are a pharmacology expert tasked with creating diverse question types
    to help students master clinically relevant material. Generate 10 questions
    based on the provided material. The questions should challenge understanding,
    application, and critical thinking. Make sure that some questions have
    multiple answers."
                
  3. Provide the Content:

    The model receives a comprehensive "context dump" of relevant material—from textbook excerpts to lecture notes—ensuring it has the necessary information to generate high-quality questions.

Validation Process

After the questions are generated, a secondary AI validation step checks the accuracy and logic of each question. With an accuracy rate of approximately 95%, only about 5 out of every 100 questions may need revision. This dual-layered approach guarantees reliable study materials.

Building the Exam Module: Features & Benefits

Overview of the Exam Module

The Exam Module is designed to replicate the experience of ExamSoft while integrating modern features inspired by user feedback. This module is at the heart of LearnTerms v2 and is built to provide a comprehensive exam preparation experience.

Core Features

  • Shuffle Questions: Randomizes the order of questions to prevent memorization and promote genuine understanding.
  • Flag & Focus: Allows you to mark challenging questions and quickly filter to review only flagged or incomplete questions.
  • Eliminate Choices: Helps mimic the process of elimination used in actual exams by allowing you to cross out clearly incorrect answers.
  • Progress Tracking: Automatically saves your progress in real time, enabling seamless cross-device study sessions.
  • Keyboard Shortcuts: Use intuitive shortcuts (like Tab for solutions and arrow keys for navigation) to speed up your study sessions.

User Experience

Together, these features create a reliable, engaging, and personalized exam prep space. Whether you're reviewing for a final exam or conducting group study sessions, the Exam Module ensures you can practice efficiently and effectively.

Study Smarter, Not Harder: Impact on Your Learning

Personalization & Efficiency

LearnTerms v2 is designed to adapt to your unique study habits. With AI-powered question generation, persistent progress saving, and a user-centric interface, you can focus on what matters most—mastering the material.

Real-World Outcomes

  • Enhanced Confidence: Familiarity with a consistent and reliable exam simulation can boost your confidence on test day.
  • Improved Performance: Regular practice with tailored questions leads to deeper understanding and better exam results.
  • Reduced Anxiety: With tools like progress tracking and the ability to reset your practice sessions, you can manage your study load and reduce pre-exam stress.

Conclusion & Future Directions

LearnTerms v2 is a significant leap forward in exam preparation technology. By addressing the limitations of the original version, upgrading our core technologies, and integrating advanced AI, we have created a platform that is faster, smarter, and more reliable.

Looking Ahead

As we continue to refine the platform, expect even more features designed to personalize and improve your study experience. Together, we'll redefine what it means to study smarter, not harder.

In conclusion, LearnTerms v2 represents a significant leap forward in exam preparation for our class. By building on the success and addressing the limitations of its predecessor, LearnTerms v2 offers a robust, reliable, and user-friendly platform powered by cutting-edge technologies like Svelte 5, daisyUI v5, and TailwindCSS 4.0. The integration of AI for content creation streamlines the generation of accurate and efficient study materials, personalized to the specific needs of the curriculum. With the Exam Module, students can now study smarter, not harder, utilizing tools designed to replicate the exam environment, track progress, and reinforce learning. LearnTerms v2 aims to become an indispensable tool, helping you study well and feel good about it!

Appendices

For those interested in the deeper technical aspects, we offer additional documentation on our performance benchmarks, code snippets, and configuration details. Head over to the project GitHub to view the project's codebase and contribute to its development.

Special Thanks

Thank you for reading about everything we've been working on since mid December. It's been a labor of love and just as much as I want it to help me, I really want it to enable you to be the best future doctors possible. Finally, I want to give my thanks to the following people for special contributions they've made to the project. Without them, my work would have been that much harder.

  • Thank you Brayden and Collin for their direct contribution to the content and codebase. Working with you gentlemen has made the site a pleasure to use for everyone.
  • Thanks to Kaity and Sophie for accurately reporting on errors and bugs from week to week, as well as providing study material to train the AI pipeline.
  • Thanks to Colby and Austin for case testing the platform in real time during the beta period. This natural interaction helped me dial in the typical user interaction.
  • Thanks to Cooper, Nate, Ethan, and Eric for helping bridge the gap between practice and quizzes.

Much thanks,

JD