2026年11月15日(日)マイページ
← 記事にもどる 中学生版高校生版 Word

「印刷する」→ 送信先で「PDFに保存」を選ぶと PDF になります。解答は最後のページです。

原田英語の時事英語ジュニア|英語ニュース ワークシート(高校生レベル)

Class ____ No. ____ Name ____________________

AIが赤ちゃんの耳で言葉を覚えた カギは「夜の復習」、家の録音100万時間以上

An AI That Learns Language From a Child's Everyday Sounds

音声・ゲームはこちらから

1Researchers at Princeton University have trained AI models to learn words using only the speech that individual children heard at home. The study was published in the journal Science on Nov. 5. It found that the models learned far better when they "replayed" recent speech, much as the brain is thought to do during sleep.

2Families agreed to have their homes recorded from the time their babies were born until they were about 3 years old. Cameras and microphones captured more than a million hours of audio and video in total. The team focused on eight of the children. It built a separate model for each child.

3Each model started out like a blank slate, with no knowledge of English sounds or words. It was fed the speech the child heard, day by day and in the same order the child heard it.

4Hearing speech every day was not enough on its own. In one child's case, the model had not learned a single word by the point that matched the child's second birthday. When the researchers added a replay of recent speech after every day of the model's learning, it learned thousands of words.

5Chatbots are trained on billions of words of text, while babies start to understand speech after hearing far less. The models recognized the sounds of words, but they did not know what the words meant. The models also learned only from sound, while real babies also see the gestures and faces of the people around them.

Words 単語

trainedA1動訓練した、学習させたtaught a skill by practice again and again
individualB1形一人ひとりの、個々のabout one single person, not a group
journalB1名学術誌、専門誌a magazine where scientists share their research
capturedB1動(音・映像を)とらえた、記録したrecorded sound or pictures with a machine
separateA2形別々のnot joined together, kept apart
blank slateA1熟まっさらな状態something that has nothing on or in it yet
recentA2形最近のhappening only a short time ago
recognizedB1動(それとわかって)聞き分けた、認識したknew what something was when hearing or seeing it
chatbots外名チャットボット、会話するAIcomputer programs that talk with people by text
gesturesB1名身ぶり(gesture の複数形)moves of the hands or head that show meaning

単語のレベル:CEFR-J(『CEFR-J Wordlist Version 1.6』 東京外国語大学投野由紀夫研究室. (URL: https://www.cefr-j.org/download.html より 2026年8月ダウンロード))

1. Reading Quiz 本文の内容に合うものを選ぼう

  1. Where was the study published?

    A. In a school newspaperB. On a chatbot websiteC. In the journal ScienceD. In a children's book

  2. How many children did the team focus on?

    A. EightB. TwoC. TwelveD. Twenty

  3. How was each model given the child's speech?

    A. All at once on the first dayB. Only on the child's birthdaysC. Day by day, in the same order the child heard itD. In a random order chosen by a computer

  4. What happened when the researchers added a replay of recent speech?

    A. The model stopped learning.B. The model learned thousands of words.C. The model learned what words mean.D. The model began to talk like an adult.

  5. Which is a difference between the models and real babies?

    A. Real babies need billions of words.B. The models learned only from sound.C. The models used gestures and faces.D. Real babies never hear speech at home.

2. True or False 正しければ T、違っていれば F

  1. (     ) The recordings continued until the children were about 3 years old.

  2. (     ) Each model already knew many English words at the start.

  3. (     ) Before the replay was added, one child's model had learned no words by the point that matched the child's second birthday.

  4. (     ) The models understood what the words meant.

3. Fill in the Blanks 下の語から選んで空らんをうめよう(形を変えずに)

separaterecentjournalrecognizedtrainedcaptured

  1. The models _____ the sounds of words, but they did not know what the words meant.

  2. Cameras and microphones _____ more than a million hours of audio and video in total.

  3. It found that the models learned far better when they "replayed" _____ speech, much as the brain is thought to do during sleep.

  4. The study was published in the _____ Science on Nov. 5.

  5. It built a _____ model for each child.

  6. Researchers at Princeton University have _____ AI models to learn words using only the speech that individual children heard at home.

4. Word Order 語を並べかえて正しい文にしよう

  1. ( children / The / on / the / of / focused / team / eight ) .

  2. ( on / was / every / speech / own / day / enough / its / Hearing / not ) .

  3. ( Science / 5 / The / published / the / in / journal / on / was / study / Nov. ) .

5. Think and Write 英語で答えてみよう

  1. The AI learned better when it reviewed recent sounds. How could you use this idea when you study English?(AIは最近の音を復習するとよく学べました。英語を勉強するとき、この考えをどう使えますか?)

  2. Do you think AI will someday learn language in the same way as humans? Why or why not?(AIはいつか人間と同じやり方で言葉を学ぶようになると思いますか?それはなぜ?)

Answers 解答

1. (1) C (2) A (3) C (4) B (5) B

2. (1) T (2) F (3) T (4) F

3. (1) recognized (2) captured (3) recent (4) journal (5) separate (6) trained

4.

  1. The team focused on eight of the children.
  2. Hearing speech every day was not enough on its own.
  3. The study was published in the journal Science on Nov. 5.

全文訳

1プリンストン大学の研究者たちが、一人ひとりの子どもが家で聞いた話し声だけを使って、単語を覚えるAIのモデルを訓練しました。この研究は、11月5日に学術誌『サイエンス』に発表されました。研究では、最近聞いた話し声を「リプレイ(もう一度再生)」すると、AIのモデルがずっとよく学べることがわかりました。脳が眠っているあいだにしていると考えられていることに似ています。

2家族たちは、赤ちゃんが生まれたときから約3歳になるまで、家の中を録音することに同意しました。カメラとマイクは、全部で100万時間をこえる音声と映像を記録しました。チームは、そのうち8人の子どもにしぼりました。そして、一人ひとりに別々のモデルを作りました。

3どのモデルも、英語の音や単語を何も知らない、まっさらな状態から始まりました。モデルには、その子が聞いた話し声が、1日ずつ、その子が聞いたのと同じ順番であたえられました。

4毎日話し声を聞くだけでは、それだけでは足りませんでした。ある子どもの場合、モデルは、その子の2歳の誕生日にあたる時点でも、単語を1つも覚えていませんでした。研究者たちが、モデルが学ぶ毎日の終わりに最近の話し声のリプレイを加えると、モデルは何千もの単語を覚えました。

5チャットのAIは何十億語もの文で訓練されますが、赤ちゃんはそれよりずっと少ない量を聞いただけで、話し声がわかりはじめます。モデルは単語の音を聞き分けましたが、その単語が何を意味するかは知りませんでした。また、モデルは音だけから学びましたが、本物の赤ちゃんは、まわりの人の身ぶりや表情も見ています。

文法ポイント:使役の have + 目的語 + 過去分詞(~してもらう、~される)

「have + もの + 過去分詞」で「ものを~してもらう、~される」という意味。本文の Families agreed to have their homes recorded は「家族は家の中を録音してもらうことに同意した」という意味。

出典:EurekAlert!(プリンストン大学の発表)、SingularityHub、Medical Xpress ほか(記事は原田英語の時事英語ジュニアのオリジナル)|https://www.haradaeigo.com/jijieigo-junior/news/2026-11-15-ai-learns-words-like-babies/