❓ i) Why do non-English languages show symbols on the canvas?

The reading canvas is designed for English input only. When you upload a Hindi, Telugu, or other script document, the PDF text extractor (PDF.js) can sometimes fail to map Unicode glyphs correctly from the font embedded in the PDF. This is a known limitation of browser-based PDF rendering — the same PDF that looks correct in Adobe Reader may extract as garbled symbols in the browser.

Solution: For non-English documents, copy-paste the text directly into the Paste Area. This bypasses the PDF font issue and the Unicode text is loaded directly, displaying correctly. The vocabulary selection, translation, and learning features all work with any Unicode text.

The Translation panel is specifically designed for multi-language output and handles all Indian and world languages correctly via the MyMemory API.

❓ ii) Is vocabulary selection like LLM tokenization?

It is similar in concept but fundamentally different in mechanism. LLMs (like GPT or Claude) use sub-word tokenization algorithms (BPE / WordPiece) that split words into fragments — e.g., "running" → ["run", "##ning"]. This is done to handle any word ever written.

LexiRead uses whitespace + punctuation tokenization — it splits text at spaces and punctuation marks to extract complete, natural English words. It then applies stopword filtering, vowel-validity checks, and length filters to identify meaningful vocabulary words. This is closer to classical NLP word tokenization than LLM tokenization, and is intentionally kept simple so you see real, readable words — not fragments.

❓ iii) Why do some "words" look merged or strange in Auto-select?

This happens when PDF text extraction does not insert spaces correctly between words (a common PDF rendering issue). The fix in v4+ applies a vowel-pattern check and length cap to reject such tokens. If you still see strange selections, use the manual filter (5+ or 6+ letters) and click Auto-select, or just click individual words yourself.

⭐ Current Rating: 7.2 / 10

✅ What's working well

Real-time dictionaryLive translation (20+ languages)Auto-select vocabularyPDF upload + text pasteQuiz engineExport (CSV/XLS/PDF)Badge progressionGRE/TOEFL bank3 themesResizable panelsSession timerWatermark print

🔧 To improve (next versions)

Offline dictionary fallbackNon-English PDF renderingSpaced repetition (SRS) for learned wordsUser-defined word notesAudio for translationsWord frequency heatmap

🚀 Future additions (v6+)

Cloud sync (Google Drive)AI-generated context sentencesVocabulary flashcard modeReading speed trackerMulti-document comparisonMobile app version

With the improvements above, the app could reach 8.5–9.0 / 10. The core architecture is strong — it's a fully self-contained, dependency-free vocabulary tool that rivals paid apps.

🏫 About LexiRead

LexiRead is a contextual vocabulary builder designed to help students, professionals, and language learners acquire new words directly in the context of what they are reading — novels, research papers, BBC transcripts, medical texts, legal documents, and more.

The philosophy: the best time to learn a word is when you encounter it naturally in a real text, not in an isolated word list. LexiRead places the dictionary, translator, quiz engine, and word bank right beside your reading canvas so you never lose the flow of reading.

👨‍💻 Built by ProximaED

ProximaED focuses on technology-driven educational tools. LexiRead is a single-file HTML5 application — no installation, no server, no login required. Open in Chrome, Firefox, Edge, or Safari.

🔒 Privacy

All your data (learned words, basket, session) stays in your browser's localStorage. Nothing is sent to any server. The only external calls are to the free Dictionary API and MyMemory Translation API for word lookups — only the specific word you click is sent.

📜 Version History

v5 — Book library topbar dropdown, page jump, Save/JSON/AutoSave, horizontal resizers, About popup
v4 — Translation fix, paste zone, watermark print, badge progression, basket export
v3 — GRE/TOEFL panel, quiz engine, theme system, vertical resizers
v1–2 — Bible reader, core PDF reading, word intelligence