Higher Education Category Submission

965 – Adamboost – AI Learning Assistant

Adamboost is an AI-powered learning assistant designed to enhance student learning efficiency and knowledge retention. The name "Adamboost" signifies its core philosophy: "Adam," representing the human student, is "boosted" by AI, much like the AdaBoost algorithm learns by focusing on difficult instances. It addresses passive learning and information overload by transforming diverse educational materials—PDFs, videos, text documents, and DOCX files—into actionable learning tools.
The innovation lies in generating concise, AI-driven summaries and personalized quizzes from user-uploaded content. These quizzes feature configurable difficulty, question types (multiple choice, open-ended, or mixed), and AI-powered evaluation for open-ended answers, providing constructive feedback.
Technically, Adamboost utilizes a Next.js frontend and a Node.js/Express backend, integrated with OpenAI for content processing, summarization, quiz generation, and evaluation. Supabase provides database and storage.
Adamboost aims to offer students a personalized, interactive, and effective way to engage with their study materials, fostering active recall and helping them identify knowledge gaps for improved academic outcomes.

https://github.com/kametayturar/outpeer-learn

The inspiration for Adamboost arose from observing common student struggles: managing vast information, the passivity of traditional study, and difficulty accurately assessing one's understanding. Students often find it challenging to convert dense materials into effective study aids or consistently practice active recall. Creating personalized quizzes or receiving constructive feedback on open-ended questions is time-consuming and often requires external support.

Our hypothesis is that providing students with an AI-driven tool capable of (1) automatically summarizing diverse learning materials, (2) generating personalized quizzes based on these summaries, and (3) offering AI-evaluated feedback for complex open-ended questions will significantly enhance their learning process. We believe Adamboost will succeed because it directly addresses the critical needs for active learning, personalized practice, and immediate, actionable feedback. By focusing on active recall and identifying specific knowledge gaps, the tool empowers students to learn more efficiently and effectively, moving beyond rote memorization to deeper comprehension, thus making advanced learning techniques more accessible.

Not applicable

Adamboost's architecture features a Next.js frontend for user interaction and a Node.js/Express backend API. Supabase serves as the PostgreSQL database and cloud storage solution, providing a robust and scalable backend infrastructure.

Technical Workflow:
The process begins with user file uploads (PDF, video, DOCX, TXT) via the frontend. The backend, using Multer for file handling, stores these files in Supabase Storage. It then extracts text content using specialized libraries: pdf-parse for PDFs, mammoth for DOCX files, and fluent-ffmpeg to extract audio from videos, which is then transcribed using OpenAI Whisper. This extracted text is subsequently sent to the OpenAI API (GPT models) for generating concise summaries. Based on these summaries and user-defined parameters (difficulty, question type, number), personalized quizzes, including open-ended questions, are generated through further OpenAI API calls. For open-ended quiz answers, the student's response is sent to the OpenAI API, which evaluates it against expected concepts and provides constructive feedback.

Innovative Feature Implementation: The AI evaluation of open-ended questions is a key innovation, achieved through sophisticated prompt engineering to guide the OpenAI model in assessing answers for conceptual understanding and providing relevant feedback. Video content processing, involving audio extraction and subsequent transcription, enables the generation of summaries and quizzes from spoken educational content.
Design & Development Timeline: Core features are fully implemented. The current phase focuses on refinement, user testing, and addressing minor bugs, with ongoing iterative improvements planned based on user feedback.

Performance Metrics:
Key metrics include:
Quiz generation time (target: under 30 seconds for average material).
Qualitative accuracy and relevance of AI evaluation for open-ended questions.
System response time for uploads and content retrieval.
User engagement metrics (e.g., session duration, quiz completion rates).

Function Points, Technical Applications, and Progress:

File Ingestion & Text Extraction:
Technical Application: Utilizes Multer for file handling, pdf-parse for PDFs, mammoth for DOCX files, and fluent-ffmpeg combined with OpenAI Whisper for video audio extraction and transcription.
Progress: Implemented.

Content Summarization:
Technical Application: Leverages OpenAI API (GPT models).
Progress: Implemented.

Personalized Quiz Generation:
Technical Application: Employs OpenAI API (GPT models) with custom prompting logic to tailor quizzes.
Progress: Implemented.

AI Open-Ended Answer Evaluation:
Technical Application: Uses OpenAI API (GPT models) guided by rubric-style prompting for detailed feedback.
Progress: Implemented.

Frontend User Interface & Experience (UI/UX):
Technical Application: Built with Next.js, Tailwind CSS, and DaisyUI; React Markdown for rendering content.
Progress: Implemented.

Backend API & Database Interaction:
Technical Application: Developed using Node.js and Express, with Supabase (PostgreSQL, Storage) for data management.
Progress: Implemented.

Adamboost’s innovation stems from its holistic and personalized AI-assisted learning approach, embodying its core philosophy of augmenting the human learner with AI's boosting capabilities.

Key creative elements differentiating Adamboost include:
User-Centric Content Processing: It processes the student's own diverse learning materials (PDFs, videos, DOCX, text files), ensuring direct relevance to their specific courses.

AI-Evaluated Open-Ended Questions: A central innovation is the generation and, crucially, the AI-powered evaluation of open-ended questions. This promotes deeper critical thinking and offers nuanced, constructive feedback that mimics a tutor's guidance, helping students understand why their answer is correct or how it can be improved.

Integrated Learning Workflow: The platform provides a seamless flow from content upload to summary generation and then to tailored quiz creation, encouraging consistent engagement with active recall principles.

Personalized Quiz Configuration: Users can adapt quiz difficulty, length, and question types (MCQ, open-ended, or mixed), tailoring the learning experience to their current understanding and study goals.

These features transform passive content consumption into an active, reflective experience. The AI feedback on complex questions is particularly innovative, addressing a significant gap in self-study by providing insights typically hard to obtain without human intervention, thus fostering more robust comprehension.

Adamboost leverages cloud-native technologies for scalability. The Node.js backend can be containerized and deployed on scalable infrastructure (e.g., serverless functions). Supabase offers inherent scalability for database and storage. OpenAI API calls are managed by OpenAI's robust infrastructure. We plan asynchronous processing for computationally intensive tasks like video transcription and large document summarization using message queues to improve responsiveness and prevent API timeouts. Caching frequently accessed summaries or question sets will also reduce redundant processing.

Environmentally, using efficient cloud providers and optimizing AI API calls (e.g., by refining prompts for conciseness, batching requests) minimizes computational overhead. As a digital solution, it inherently reduces paper consumption.

Long-term user engagement will be fostered through continuous improvement based on user feedback, introducing new features like progress tracking, spaced repetition algorithms, and potentially collaborative elements. The system is designed modularly, allowing for easy integration of newer AI models or educational tools. Adapting to evolving user needs will involve regular feedback collection and agile development sprints, ensuring Adamboost remains a relevant and valuable learning companion.

Adamboost aims to democratize access to personalized and effective learning tools, addressing educational equity. By providing an AI-powered assistant that adapts to individual learning materials (like PDFs, videos, and documents) and styles, it enhances the lives of its primary beneficiaries—students—by making study more efficient, engaging, and impactful. This can particularly benefit students who lack access to private tutoring or extensive academic support. The AI-evaluation of open-ended questions provides a form of personalized feedback crucial for developing critical thinking.
The project aligns with broader social goals such as inclusion by supporting diverse material formats and offering customizable learning experiences. It can help bridge achievement gaps by providing students with tools to master complex subjects at their own pace. Metrics to measure social impact will include user adoption rates, particularly among students from diverse backgrounds; surveys assessing perceived improvements in learning outcomes and confidence; and feedback on the quality of AI-generated content.
We will ensure responsiveness to community needs through active feedback channels and iterative updates. Ethical AI use is paramount: we will strive to mitigate biases in AI-generated content and ensure data privacy and security for all users, fostering an equitable educational landscape.

Personal Information Collection Statement (PICS):
1. The personal data collected in this form will be used for activity-organizing, record keeping and reporting only. The collected personal data will be purged within 6 years after the event.
2. Please note that it is obligatory to provide the personal data required.
3. Your personal data collected will be kept by the LTTC and will not be transferred to outside parties.
4. You have the right to request access to and correction of information held by us about you. If you wish to access or correct your personal data, please contact our staff at lttc@eduhk.hk.
5. The University’s Privacy Policy Statement can be access at https://www.eduhk.hk/en/privacy-policy.

• I have read and agree to the competition rules and privacy policy.