Tomo-Sensei is the Japanese grammar AI explainer built into Yomimaru's reading interface. Highlight any sentence, in a library text, in something you pasted in, or in a document you uploaded yourself (a novel as an EPUB, a PDF), and you get a structural breakdown of that exact sentence: what each particle is doing, where the embedded clauses begin and end, who the dropped subject is, and what the conjugation chain means. A grammar reference explains the pattern. This explains your sentence.
The gap that grammar books cannot fill
Every serious learner owns a grammar reference: Genki, Tae Kim, A Dictionary of Japanese Grammar, Bunpro. They are good at what they are for: telling you what te-form is, what a passive looks like alone. The gap opens when you meet a real sentence, recognise every pattern in it, and still cannot say what it means.
長年その問題を避けてきたとされる政府の対応が、ようやく見直されつつあるとの報告が出た。
Read it once. Now say who has been avoiding the problem, and where the main clause starts. Every word is one an N2 reader knows. The sentence is still a wall.
What stopped you is the shape. Dictionaries describe patterns. None will take this sentence and tell you how its patterns are stacked. That is the job Tomo-Sensei does.
What actually makes a Japanese sentence hard to parse
Two features do most of the damage, and neither is a grammar point you can look up.
The verb arrives last. 「ケンはりんごを食べた」 gives you Ken, then the apple, and only then 食べた. Across five words that costs nothing. Across four lines it costs plenty: you hold every noun and particle in memory without knowing what they are for, because the word that assigns their roles has not arrived. Lose one particle on the way and you cannot get it back.
Modifiers come before the noun, and nothing marks where they start. English announces the description with a joining word: the book that my friend read yesterday. Japanese puts the clause in front and announces nothing.
昨日友達が読んだ本
直訳 Word by word
Yesterday friend read book.
Four words, and only the last is the thing being talked about. No "that", no comma, no signal that a clause has opened. You find out when you hit 本.
Reading a long pre-nominal clause as if it were the main sentence. You see 昨日友達が読んだ, you find a subject and a past-tense verb, you close the sentence in your head, and then 本 turns up and nothing fits. Nest two of these and you go back to the top twice before working out that neither verb was the main one.
That is what Tomo-Sensei is built for.
What Tomo-Sensei actually tells you
Particle disambiguation in context
は marks the topic, が the subject, を the object. That stops being enough the moment a sentence carries two verbs: the question is no longer what が means, but which verb this が belongs to.
私が好きな人は日本にいます。
直訳 Word by word
I (subject) like person, as for, in Japan is.
That が has nothing to do with います. It belongs to 好きな, inside the clause describing 人. Misread it as the main subject and you are the one in Japan.
Embedded relative clause structure
One noun can carry three stacked clauses in front of it, each with its own verb and particles. Alone, each is recognisable. Stacked, they are not, and that is where intermediate readers stall. Here is the sentence from the top, in pieces.
長年その問題を避けてきたとされる政府
The whole first half is one modifier. It is not making a claim about a government. It is naming one.
政府の対応が、ようやく見直されつつある
Now the が. What is under review is 対応, not the government you spent the first half reading about. 見直されつつある is passive plus 〜つつある: under way, no agent named.
〜との報告が出た。
And there is the main clause, four words, at the very end. Everything before it was a quotation carried by と. The sentence is about a report.
Dropped subjects and implicit speakers
Japanese drops the subject whenever context can supply it, and in a novel the context can be two pages back.
部屋に入ると、もう帰っていた。
直訳 Word by word
Room-into entered-when, already had-gone-home.
Two verbs, two different people, neither named. You cannot write the English without deciding who they are. Tomo-Sensei reads the surrounding text and tells you who each is.
Passive and causative constructions
Passive (〜れる/〜られる), causative (〜せる/〜させる) and causative-passive (〜せられる/〜させられる) stack onto the tail of a verb and rearrange who is doing what to whom. Newspapers compress them, and at speed they mislead.
部長に残業させられた。
直訳 Word by word
By the department head, made-to-do overtime, was.
Causative-passive. That に is not a destination. It marks the person who made it happen. Two kana, させ, and the agency moves from you to him.
Conditional nuances
〜たら, 〜ば, 〜と, 〜なら. The books explain the difference in theory, the only place it is ever clean. Tomo-Sensei names which one is in your sentence and what it implies there: a neutral if-then, a hypothetical, an expectation, a recommendation.
Formal written register patterns
N2 and N1 prose runs on written patterns almost nobody says out loud. They look like three or four words and behave like one.
- 〜にすぎない
- is no more than, is merely
- 〜にもかかわらず
- in spite of, despite
- 〜をめぐって
- over, concerning (a dispute or a debate)
- 〜に基づいて
- on the basis of, according to
Parse 〜にすぎない piece by piece (に, すぎる, ない) and you get "does not exceed", which is neither English nor the point. Tomo-Sensei reads them whole.
How Tomo-Sensei differs from other grammar tools
| Grammar dictionaries | Bunpro | DeepL / Google Translate | ChatGPT (standalone) | Tomo-Sensei | |
|---|---|---|---|---|---|
| Works on a specific sentence | ✗ | ✗ | Translation only | ✓ | ✓ |
| Integrated into reading interface | ✗ | ✗ | ✗ | ✗ | ✓ |
| Works on text you bring yourself (pasted or uploaded) | N/A | N/A | ✓ | ✓ | ✓ |
| Preserves reading flow | ✗ | ✗ | ✗ | ✗ | ✓ (inline) |
| Context-aware (reads surrounding text) | ✗ | ✗ | ✗ | Sometimes | ✓ |
The last two rows matter most. Every other tool on that table asks you to leave the page: copy the sentence out, paste it elsewhere, read the answer, come back, find your place. Four context switches for one sentence you were already struggling to hold.
How to use Tomo-Sensei effectively
Reach for it when a sentence is a wall: read twice, every word known, clauses refusing to connect. Vocabulary is not its job: tap the word instead.
- Read the sentence. If the meaning comes, keep going.
- If it blocks you, parse it yourself first: the main verb at the end, then the bare subject-object-verb frame under the modifiers.
- Highlight it and compare the breakdown against your attempt. The gap is the lesson.
- Re-read the Japanese with the right structure in your head. That is the pass that sticks.
By JLPT level: what Tomo-Sensei helps with most
| Level | What starts blocking you |
|---|---|
| N4 and N3 | Embedded relative clauses, dropped subjects, te-form chains, the first passives. |
| N2 | Stacked modifying clauses in formal texts, hedging and perspective markers, compound conditionals, vocabulary that works as one unit. |
| N1 | Arguments layered across paragraphs, honorific and humble chains, academic hedging, sentence-final nuance particles. |
Tomo-Sensei as a grammar pattern database
Every explanation is tied to the sentence you highlighted, so across an hour of reading you meet the same construction three or four times and see it explained in each. A grammar list tells you what a pattern is. Only reading native Japanese content shows what it does.
The point of the breakdown is that you stop needing it. A sentence you parse with help this week is one you parse unaided next month, because you watched the pattern work in a real sentence rather than read about it in an entry.
Further Reading
- JLPT N3 reading comprehension strategies. Grammar knowledge applied to timed reading passages at N3.
- JLPT N2 reading strategies. The structures and formal register N2 reading throws at you.
- Master Japanese vocabulary through context. Vocabulary and grammar reinforce each other when you build both by reading.
Test yourself: the adaptive grammar challenge
The challenge below pulls a particle or a conjugation out of a real sentence and asks you to put it back: five questions, adapting to the level you clear. Spotting clause structure under pressure is a different skill from reading about it, and guessing on the embedded-clause items is the gap Tomo-Sensei closes while you read.
