How Todoku works

Comprehension scoring

When you upload a subtitle file or paste lyrics, Todoku tokenizes the text using kuromoji (a Japanese morphological analyzer) to extract content words: nouns, verbs, adjectives, and adverbs. Function words like particles (は, が, を) and auxiliaries are excluded since they don't carry meaning on their own.

Unique words are counted in dictionary form, so 食べた and 食べる count as one word. Frequency is tracked so you can prioritize the most common unknown words first.

JLPT word lists

The JLPT (Japanese Language Proficiency Test) organizes vocabulary into five levels, N5 (beginner) through N1 (advanced). Todoku uses curated word sets for each level as the baseline for what you should know at your placement level.

These lists are based on commonly referenced JLPT vocabulary resources. They are not official. The JLPT does not publish an official word list, so treat them as a reasonable approximation, not a definitive standard.

The lists are cumulative: if you're placed at N3, Todoku assumes you know all N5, N4, and N3 words. You can also mark individual words as known to fine-tune your score.

Placement test

The placement test shows 20 words (5 per level from N5 to N2) and asks if you recognize each one. Readings are hidden by default. The goal is kanji recognition, not phonetic reading. Your level is inferred from how many words you know at each level. You can skip the test and set your level manually if you already know where you are.

Why I built this

I've been watching anime for 15+ years. Somewhere along the way I developed a genuine interest in learning Japanese and tried a lot of approaches: apps, textbooks, structured study plans. But no matter what I tried, I always ended up back where I started, watching anime and listening to J-pop.

Eventually I stopped fighting it. If that's how I naturally spend my time anyway, why not turn it into the study? Todoku is the tool I wished existed: something that meets you where you already are, and tells you exactly how much you understand and what to learn next.

Limitations

• • Scoring is based on vocabulary recognition, not grammar comprehension. You could score 90% and still not understand a sentence due to unfamiliar grammar patterns.
• • The JLPT word lists are approximations and may differ from other resources.
• • Subtitle files vary in quality. Machine-generated or poorly formatted files may skew results.
• • Named entities (character names, place names) are counted as unknown words even if you'd recognize them in context.