Audio-Only Lecture Learning App

ISEF Category: Systems Software

Subcategory: Online Learning  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A weak internet connection can turn a 1-hour lecture into a dead end. If your app can shrink that lecture into something a phone can handle, you could help students who live far from stable broadband. The hard part is not just saving data. You also need to prove that people still learn from it.

What Is It?

This project asks a simple question with a tricky answer. Can you compress a lecture so far that it works on very slow connections, while still helping students understand the content? Your idea uses two pieces. First, it turns speech into very small audio files. Second, it extracts chapter slides from the video with OCR, which means optical character recognition, the software that reads text from images.

Think of it like making a travel-sized version of a textbook lecture. You keep the voice, which carries tone and timing, and you keep the key slides, which act like signposts. The challenge is balance. Cut too much, and learners lose context. Keep too much, and the app stops being useful for rural areas with low bandwidth.

Why This Is a Good Topic

This is a strong science fair topic because you can test it with real users, real data, and clear outcome measures. You can compare your format against a common baseline like low-resolution video and measure comprehension, recall, completion time, and data use. The project also connects to a real access problem, which gives it clear social value. You can learn app design, information compression, OCR, and experiment design without needing a full hardware lab.

Research Questions

• How does an audio-only lecture format with OCR slides affect quiz scores compared with low-resolution video?
• What is the effect of slide density on comprehension when learners use an audio-first lecture app?
• Does chapter segmentation improve recall more than a continuous audio track does?
• To what extent does transcript availability change completion rates and note quality?
• Which lecture features, such as diagrams, formulas, or dense text, cause the largest drop in understanding after compression?
• How does playback speed choice affect comprehension in a low-bandwidth lecture app?

Basic Materials

• Smartphone or laptop with audio playback capability.
• Headphones.
• Sample lecture videos with clear slide changes.
• OCR-capable text extraction tool.
• Simple quiz platform or form tool.
• Spreadsheet software for data logging.
• Screen recording or usage tracking tool.
• Stable internet connection for setup and testing.

Advanced Materials

• Server or workstation for media processing.
• Audio compression and transcoding software.
• OCR pipeline software.
• Lecture video dataset with permission for research use.
• Pretest and posttest survey platform.
• Statistical analysis software.
• Database for storing chapter metadata and quiz results.
• Accessibility testing tools for captions and transcripts.

Software & Tools

• Python: Automates video processing, OCR runs, and data analysis for the app prototype.
• FFmpeg: Converts lecture video into compressed audio and extracts media segments.
• Tesseract OCR: Reads text from slide images and turns it into searchable chapter labels.
• ImageJ: Helps inspect slide image quality and compare extracted frames.
• R: Tests whether comprehension differences are statistically meaningful.

Experiment Steps

1. Define the learning task you will measure, such as recall, comprehension, or completion rate.
2. Choose the baseline format you will compare against, such as low-resolution video, audio plus transcript, or audio plus slides.
3. Decide how you will split lectures into chapters and which slide features OCR must capture.
4. Build a scoring plan that links media format to learning outcomes and data use.
5. Plan controls that keep lecture difficulty, topic, and test timing as similar as possible across groups.
6. Pre-register the analysis you will run, including the main comparison and any subgroup checks.

Common Pitfalls

• Measuring only data size and forgetting to test whether learners actually understood the lecture.
• Using lectures with wildly different topics, which makes format effects impossible to separate from content difficulty.
• Letting OCR miss small text or diagrams, which breaks the chapter slide logic and confuses users.
• Comparing your app to an unfair baseline, such as blurry video with no captions or navigation.
• Collecting too few participants, which leaves the comprehension results too noisy to trust.

What Makes This Competitive

A stronger version of this project does more than build a demo. It tests a clear hypothesis with a careful study design and enough participants to support a real claim. You can raise the level by comparing several compression and navigation strategies, not just one. Strong projects also report tradeoffs, such as when audio-first learning helps, when it fails, and which lecture types suffer most.

Project Variations

• Test the same app with STEM lectures versus humanities lectures to see whether slide OCR helps one subject more.
• Compare OCR-generated chapter labels with human-written chapter labels to measure whether automation changes learning quality.
• Add multilingual captions or translated slide text and measure whether the system helps bilingual learners more than monolingual learners.

Learn More

• PubMed: Search for review articles on multimedia learning, cognitive load, and online education outcomes.
• NIH National Library of Medicine: Search for open-access studies on learning, attention, and comprehension measurement.
• NASA eClips: Explore free examples of short-form educational video design and chunked lesson structure.
• MIT OpenCourseWare: Look for free courses on machine learning, computer vision, and information retrieval.
• International Journal of Artificial Intelligence in Education: Search for peer-reviewed work on adaptive learning systems and learner assessment.