Portion Cut: paste your exam portion, read only the pages that matter (EmbeddingGemma, in the browser)

Описание и применение

Portion Cut автоматически сопоставляет строки экзаменационного билета с страницами учебника, выделяя только нужные страницы и создавая компактный PDF, что экономит время студентов.

Задача

Students waste time searching textbook pages that match exam syllabus lines, especially when wording differs and topics span multiple chapters.

Результат

On a 397‑page textbook, Portion Cut identified 99 pages to read for the tuning portion and 75 for the held‑out portion, found 18 of 21 and 12 of 16 needed sections respectively, and correctly flagged all missing lines without false positives.

Рабочий процесс

• Paste exam portion text into the web app.
• Drop the textbook PDF into the app.
• The app extracts text from each page using pdf.js.
• Each page is embedded with EmbeddingGemma 300M (4-bit ONNX) and cached in IndexedDB.
• Each syllabus line is embedded and matched to page vectors; missing lines are flagged based on score thresholds.
• Relevant pages are grouped by section and assembled into a smaller PDF with pdf-lib.
• The resulting PDF is presented for download or viewing in the browser.
• The app runs entirely client‑side with no backend or API key.
• The demo mode uses pre‑computed vectors for a sample book.
• The tool works offline after the first visit.
• The model and embeddings are cached for instant reuse on subsequent runs.

Доказательства

Portion Cut does that matching for you. Paste the portion, drop in the textbook PDF, and it tells you which pages to read, grouped by syllabus line, with the section names next to each range.

Источник

Everything runs in the browser tab. The PDF is never uploaded, there is no backend, and there is no API key.

Источник

The matching is done by EmbeddingGemma 300M, Google's open‑weight embedding model, using the 4‑bit ONNX build from onnx‑community/embeddinggemma‑300m‑ONNX and Transformers.js on WebAssembly.

Источник

On that run, Portion Cut said to read 99 of 397 pages, found 7 of the 9 syllabus lines, and flagged the other two (sets and regular expressions) as probably not in the book.

Источник

I wanted a number I could defend, so I measured it on one free textbook, OpenStax's Introduction to Python Programming (397 pages), against two portions copied word for word from a public university Python syllabus.

Источник