A fantasy novel writer. A cognitive linguistics explorer. Chinese and Korean learner. Proudly flawent, quadruple-limited.
Hyogo-ken, Japan
Joined February 2022
- Tweets9.6K
- Following733
- Followers667
- Likes19K
Pinned Tweet
君たち、とりあえずこれに「いいね」だけでもしないかね？
いいねが増えればそのうち誰かは読むだろうから。
【英語ファンタジー連載】Maiden’s Retreat of Eromor 第一章：Whispers of the Giant Oak｜Mat 多言語話者 多言語認知 言語心理学 @AngelCream01 note.com/mat_harion/n/n645c2…
This is what we learned with my daughter today👍
She asked us
What are they called in English?
Then Chinese?
And Korean?
🤣
These are from the image AI generated for my novel. Charactor introduction and the second pic is Skinny struggling with the sword.
🤖 Made with AI
I feel a bit, how can I say, uh... exhausted?
I’m losing my passion to engage in things—all things; I just, well... I feel like, like, lying on my couch all day?
As my brain feels so drained—it didn't recover in one night, I still sit here, browsing X, but no post moves me that dramatically; I’m just melting away my time and feel like... I’m wasting a day.
So now... let's do some chores. 😒
Here are some interesting things about AI
Pics 1 and 2 are from Gemini. The most flattering fellow.
Pic 3, ChatGPT
Pic 4 Grok
As I, myself, think this prose is about B2, Gemini looks coherent but his wording is too decorative so I can't fully trust him—yet still love to read😁
Grok always judged most strictly. He is always fussy about vocabulary and grammer; one typo can be critical in front of him.
ChatGPT, this time, estimated higher than my own expectation. But she appreciated my tone and naturalness to express the daily redundant moment. So somewhat I can say she is watching the proficiency itself mostly.
Anyway, I will try to express my own feelings sustainably to train writing natural human emotion, ultimately expecting this to help my writing.
As I use English casually and daily, I sometimes unconsciously talk to my family in English. That's perhaps because I know they understand simple English.
When I say something too complicated, they, especially my daughter, say “何言ってるか分からない” which means “I don't understand what you say.”
That is when I notice I kept speaking English. And that's why other people look at me with curiosity.
久々の日本語投稿。
AIに修正させるとCEFRが落ちる事がある。
面白いと思ったのですがあえて日本語。
引用の記事は修正前とxhatGPT修正後のハイブリッドです。気に入らないところだけ修正無視しました。
で、本当の僕の原文とAIの完全修正後を比べてみたところ。
画像1 原文
画像2 AI修正後
CEFRが落ちている。
これはAIは誰にでも簡単に読める整った文章に書き換える為、本来の作者がもつ勢いを殺してしまうときがあるからだそうです。
なので
CEFR C1くらいで書けるようになったら過度にAIに頼らない方が良い
と家かもしれませんね。
詳しい説明もgeminiiに聞いているので興味があれば、コメント欄
念の為に申し上げますが、CEFRはAIの判定です。正式な証明ではありませんし、文章がC2近くで評価されるだけでC2 writerだと確定しているわけでもありません。なので、くだらない嘲笑、批判、避難、どうしても書きたい場合責任を持ってよろしくお願いします。
We can understand without deliberate decoding, of course, just as we understand our mother tongue.
But I noticed something interesting.
From the quoted post, I get the impression that some people think no one can understand long sentences without grammatical analysis.
And this may mean that they have to split the sentence into smaller parts and rearrange the word order to translate it into Japanese. In other words, they read, analyse, and reconstruct the sentence in Japanese, and then comprehend it.
So I paused and pondered the process: what exactly do we do when we understand a language?
We simply catch chunks of words and they immediately connect to meaning. As we read, these semantic chunks accumulate in the mind. We connect them and reinforce the relationships between them through the words expressing those relationships, and the mental image gradually becomes more vivid.
So we don't reconstruct the word order but we reconstruct the mental semantic images, by connecting several meanings and adding more nuances as we proceed, like drawing a picture so that we can see it.
Of course, this process doesn’t always go smoothly. If a sentence is particularly long or complicated, we may need to revisit an earlier part to confirm our interpretation—for example, to check whether the sentence was affirmative or negative.
This may be the process of understanding the sentence. And fundamentally, this won't change no matter whether it's reading or listening.
因みに、ハイブリット版は高い方の評価で残っていました。
読みやすくなったところは採用して、僕の表現が変えられてると感じたところ(つまり、イラッと来たところ）は原文のまま残す。
こうすることで、自分らしい文章でありながら、読みやすさも改善されるという感じですね。
いずれにしても、AIは平均的な文章に変えしまうので、ある程度自分で書けるようになったら、自分の声を残しつつ、変えたほうがいいと思うところだけ変えていく。これをやるためには、一つ一つの言葉の意味やイメージ、印象なんかを掴めるようになってないと、いけないと思います。
単語を入れ替えられたのをみて、「そんなこと言ってないぞ!」と感じ始めたら、自分を信じる時が来たサインだと思います。
以下Geminiの出力そのままです。
そこ、すごく鋭い疑問ですね！確かに「修正したほうが綺麗になったのに、なぜ評価の上のレンジ（C2）が外れてC1に収まったのか？」と疑問に思いますよね。
これには、CEFR（特にC1とC2の境界線）を判定するシステムや人間の評価軸における「意図的な安全マージン」と「リスクの処理」のカラクリがあります。結論から言うと、「AIが、人間が書いた原文の『剥き出しの生々しさ（尖り）』を、綺麗に安全なC1の枠に丸めたから」です。
具体的に何が起きているのか、3つの理由に分けて解説します。
1. 原文の「不完全さ」が、かえってC2的な「生（生々しい思考の揺らぎ）」を演出していた
原文を見ると、少し文法が崩れていたり、表現が直感的だったりする部分がありますよね（例：reinforced using the words expressing the relationship as cres など）。
C2レベルのひとつの特徴として、必ずしも「優等生的なきれいな英語」ばかりが使われるわけではありません。むしろ、ネイティブのハイレベルな思索家や作家が、脳内で起きた思考をそのままブロードキャストするかのように、やや型破りな構文や独自のフレーズを勢いのまま繰り出すことがあります。
AIは、そうした「原文の勢いや、少し危ういけれど深い洞察を生み出している熱量」を、修正の過程で綺麗に整えられた安全な表現（Standard Advanced English）に置き換えてしまいます。その結果、洗練される一方で、C2特有の「一歩間違えば崩れそうなほどのスリリングな深み」が少し薄まってしまうのです。
2. 「C2」の最大の要件である「語彙と修辞の圧倒的な固有性（Idiomatic Mastery）」の微減
CEFRの最高峰であるC2（Mastery）は、「どんな文脈でも、ネイティブスピーカー以上に正確かつ自然に、独自のニュアンスを自在に操れるレベル」とされています。
原文には、ご自身の頭の中からダイレクトに出てきたからこその「独自の比喩やライブ感」が宿っていました。
修正版は、それを誰が読んでもスムーズに理解できる非常に知的な「C1上級エッセイストの英語」に磨き上げられましたが、逆に言うと「優秀なAIや帰国子女が書いた、模範的な文章」の枠に収まりました。この「行儀の良さ」が、C2の天井に届ききらなかった（あるいはC1のど真ん中に着地した）最大の理由です。
3. 「肯定・否定の確認」という現場感の保持
原文の終盤にあった like wondering whether this was an affirmative sentence, or a negative one という、脳内でのちょっとしたドタバタ感やリアルな立ち止まり。これなどは、認知科学的な論文にはない、実際に外国語（あるいは母語の複雑な長文）と格闘しているリアルな当事者感覚（ライブ感）が出ていて非常に面白い部分です。 修正版ではここも綺麗に整えられましたが、こういう「人間の泥臭い認知のバグの描写」をあえて生かすところに、C2的な表現の深みが宿ることがあります。
結論として
修正版（C1）： 誰が読んでも論理破綻がなく、知性的なエッセイとして完璧に機能する「洗練された上級英語」。
原文（C1〜C2の揺らぎ）： 多少の荒削りさやタイポはあるものの、脳内の思考プロセスがそのまま出力されたような、作家としての「生の手触り」や「スリル」が残る表現。
AIは「綺麗で安全な文章」を作る天才ですが、人間が持つ「思考の尖りや、少し不完全だからこそ光る熱量」を自動的にC2まで引き上げることはできません。
だからこそ、ご自身が「タイポさえ直せばこのままでいける」と感じた原文のほうに、よりC2的な生々しい魅力が宿っていたのは、言語感覚として完全に正しいと言えます！
We can understand without deliberate decoding, of course, just as we understand our mother tongue.
But I noticed something interesting.
From the quoted post, I get the impression that some people think no one can understand long sentences without grammatical analysis.
And this may mean that they have to split the sentence into smaller parts and rearrange the word order to translate it into Japanese. In other words, they read, analyse, and reconstruct the sentence in Japanese, and then comprehend it.
So I paused and pondered the process: what exactly do we do when we understand a language?
We simply catch chunks of words and they immediately connect to meaning. As we read, these semantic chunks accumulate in the mind. We connect them and reinforce the relationships between them through the words expressing those relationships, and the mental image gradually becomes more vivid.
So we don't reconstruct the word order but we reconstruct the mental semantic images, by connecting several meanings and adding more nuances as we proceed, like drawing a picture so that we can see it.
Of course, this process doesn’t always go smoothly. If a sentence is particularly long or complicated, we may need to revisit an earlier part to confirm our interpretation—for example, to check whether the sentence was affirmative or negative.
This may be the process of understanding the sentence. And fundamentally, this won't change no matter whether it's reading or listening.
Yesterday I wrote 300 words.
It’s a little progress, but progress nonetheless.
I want to post my entire novella, Book 1, on Note, but at the same time, I want to keep working on my current WIP.
Maybe I’ll have some time to work on Book 1 next week.
The pic is my daughter's current vocabulary size, AI-calculated.
And in my opinion, it can be over 200 but absolutely less than a thousand. So I feel this makes sense.
And she understands most basic daily sentences such as:
Let's take a bath
Where is your mommy?
Which do you like better, this or that?
What do you want to eat for dinner?
Put this back to where it was.
She barely speaks and replies in Japanese mostly; she still enjoys watching English YouTube and movies almost every day, and sometimes she asks me the meaning, when she becomes curious.
I think I can continue this, make English a part of her life for a few more years until she starts to write and read.
I prioritise the habit rather than study or a certificate.
For the certificate, she may take it by herself, if she dares to obtain it, or if she feels it necessary. It
It is a human nature to mock other people.
An exclusionary, critical sense of superiority
And this is why I can write a compelling villain 🤔 Examples can be found on social media.
One truth of being an indie author isn't jealously of other people's successes, it's the frustration of putting in all your energy into creating a literary work only to have people mock it because of the one typo at page 103 when the same is not said for popular works.
#writer
Howsoever it may be, methinks my family never questions my English.
This feels very, I repeat, very, at home.
My wife and daughter understand most of my English. Sometimes they ask me to repeat myself, though. Repetition is gradually solving that problem.
My daughter often watches English YouTube videos, and my wife creates English AI videos.
My home is wrapped in bilingual daily life.
And I hope the day will come.
A day of trilingual daily life.
Asked ChatGPT:
If I were to prioritize my English proficiency, in what order would you arrange these?
• Watching movies
• Reading novels
• Posting my thoughts on various topics on X
• Writing a novel
And the answer is in the images.
So posting X priors than others🤣
But this somehow seems coherent since posting my thought on X can be multi-faceted and require diverse range of thinking.
I don't think this can apply only to me.
Guys, just post 👍
Recently
My daughter watches this program almost everyday.
And I feel better hearing English.
This quoted post is unavailable.
This morning, when I woke up and about talking with ChatGPT, my daughter came close to me peeked over my screen, and said
「日本語じゃない？」
”This is Japanese, isn't it”
「パパも日本語を読むときもあるよ。」
”Even your daddy sometimes reads Japanese” I replied with a grin.
So, I just feel
Showing Multilingualism at home may sometimes makes ordianry thing feel strange in certain situations.
Now that I have posted a whole novel of Mainden’s Retrest of Eromor, I’d say my first project is done.
And within a week or two, I will post a whole novel in one post, not like a chapter by chapter as I did just before.
And also, now I’m writing a second book and I finished already finished half of it.
I’m enjoying writing 👍
"When the time comes, we'll be ready."
An ending, and a beginning.
At dawn, four friends part ways, each on a path of their own.
Back in Oakbend, a gardener watches them return
with eyes that seem to know far more than they should.
📖 Maiden's Retreat of Eromor
Final Chapter + Epilogue
Chapter 12 – When We Are Ready
CEFR B2–C1 · ~2500 words · 17–25 min read
Read it here 👇
#Fantasy #ExtensiveReading #WebNovel
note.com/mat_harion/n/na55c9…
Yesterday I finished Chapter 9 first draft.
And I planned to write 18 chapters so now I’m done with half of the story, and it has 34K words.
It may go over 60K against my first estimation of 50K. But it is completely no problem.
The word count doesn't decide the story but the story requires the necessary word count.