Bookmark Learning Center (Website/IOS/ANDROID)

A Model for Word-Learning Flow

I think Language Reactor (LR) provides an excellent model for how a “flow” of adding words to learn could work. https://www.languagereactor.com/phrasepump

When watching a video—say on Netflix—you can simply click a word you want to learn. Behind the scenes, LR automatically saves a screenshot and captures the original audio from the source. It feels seamless and effortless.

Because LR relies on a frequency list, it suggests high-frequency words first. This helps learners prioritize vocabulary that matters most for comprehension. To make this clear, LR color-codes words: orange for suggested words, and purple for words not yet suggested. Once you’ve worked through the higher-frequency words, those purple ones eventually move into the suggested list.

PhrasePump Overview

LR also offers an SRS system called PhrasePump, which is built almost entirely around Cloze-style cards. At first, I disliked this approach, but after experimenting with it, I developed a method that completely changed my perspective—and now I find it incredibly effective.

What is PhrasePump?
PhrasePump is an SRS feature that works with full sentences instead of isolated words. Instead of traditional flashcards, it presents Cloze-style prompts where the target word is hidden in context. Learners review these by recalling the missing word or listening to the full audio to strengthen comprehension. Importantly, PhrasePump prioritizes vocabulary using a frequency list, ensuring focus on the most impactful words.

My Use Case

As an intermediate learner of Japanese, I reset all of my Known and Learning words in LR and began using PhrasePump with a specific strategy:

• Focus on comprehensible input. I target sentences where I already know about 98% of the words, leaving only 1–2 unknowns.

• Use Cloze recall for unknown words. This engrains the word in context and builds retrieval strength.

• Use audio playback for reinforcement. If I already know the word, I practice listening to ensure I can catch it in natural speech.

• Maintain control. I decide when a word is “Known.” If it doesn’t stick, I delete it and wait until it naturally reappears in immersion.

It also has a “Suggestion” Mode where it can fill gaps from immersion and suggest pre-made sentences with audio that use those high frequency words that may have been missed.

Why This Works

• Keeps study tied directly to immersion content.

• Frequency list ensures focus on high-value vocabulary.

• Combines Cloze recall with listening practice.

• Gives the learner full control over what to keep, repeat, or drop.

Suggestion for FluentAI

A system like this would give learners natural, sentence-based exposure while still leveraging SRS. If you combine frequency-based word targeting with learner control (marking, deleting, or reinforcing items at will), you create a system that avoids the rigidity of traditional flashcards while keeping the benefits of spaced repetition.

More Control

I’ve come to really appreciate Cloze-style cards, but they’re not for everyone. Some learners prefer a more automated system that decides when a card is “unknown,” “learning,” or “learned.” Personally, I’d like more advanced options. For example:

• When I mark a card as Known, I don’t want to see it again in SRS.

• If the same word later appears in a new context I don’t fully understand, I’d like to re-add it as a fresh card—without building up an endless review queue.

Most language learning apps only track words in a single dimension, but listening and reading are distinct skills. Designing a system that tracks a word across recall, listening, and reading could open the door to smarter ways of progressing vocabulary through an AI-driven SRS.

Thats amazing, but iam still observed this apps.

The app is great, the best I've ever used for learning, and I hope it can create subtitles using AI

nice