弁論ブログ

ディベートブログですがニッチなことを書きます。

Youtubeの英語字幕生成機能がけっこうすごい

こんにちは。

 

しばらくサボっていましたが(これは特に理由があるわけではなく、ただサボっていただけです。)、年も明けたので更新します。

 

 

Youtubeには自動で動画の字幕を生成してくれる機能があります。

今日は、これの英語版の精度がけっこうすごくて、便利という話をします。

 

ちなみに日本語の自動字幕生成はまだまだイマイチで、

それをネタにしたヒカキン自動生成字幕botアカウントみたいなやつもあります。

f:id:kief_forensics:20190107185411j:plain

 (このヒカキンは岡橋に似ている)

 

 

(1)どのくらいの精度か

 WUDC 2016 FinalのPMを聞いてみましょう

動画再生画面の右下の字幕ってボタンおよび、Youtubeページ上のいいねボタンの隣の・・・ってタブの文字起こしを開くってところから自動生成字幕が読めます。

 

 

 

Youtubeが自動生成した字幕

madam chair the global poor all around
the world and no matter what country in
which they live currently live in a
system of dictatorship they live under a
dictatorship known as no alternatives
shackled by capital that's been unjustly
acquired constrained by landed gentry
who have no incentives but to pursue
their own interests and chained by the
fact that they can't do anything but to
look at the question of their own
subsistence they're unable to reach out
for the right to Liberty and to
self-determination that we think in
here's in the human condition

 

比較のためにボセオファンページの(人間が書き起こした)スクリプトも載せます。

Madame Chair, the global poor all around the world, and no matter what country in which they live, currently live in a system of dictatorship. They live under a dictatorship known as no alternatives, shackled by capital that’s been unjustly acquired, constrained by landed gentry that have no incentives but to pursue their own interests, and chained by the fact that they can’t do anything but to look at the question of their own subsistence. They’re unable to reach out for the right to liberty and to self-determination that we think inheres in the human condition.

 

完璧ですね。

 

 

もうちょっとスピードが早いスピーカーでも結構いけます。

EUDC 2014 FinalのMOを聞いてみましょう

 

mr. speaker the problem with the
feminist movement isn't that they're not
they're saying the wrong things
it's that nobody's listening we think
that if we can prove in this debate that
you harm the ability of the feminist
movement to achieve its aims at home
then on the comparative this isn't a
world in which we want to live in
because the comparative in this debate
has been completely missed right and
this is why I came is not to the looking
baffled because the problem in the
Western liberal democracies a moment is
that feminist movements are small and
peripheral as actors that low numbers of
women and even lower numbers of men self
identify as feminists people in this
room I self identify as feminists but
that's because you're spending a
Saturday evening listening to our
history on the Grieg of speech in the
national theatre right so on the
comparative these conflicts happen
anyway the feminist movements do not
have the ability to stop these walls
occurring no conflict has ever looked
unlikely until the feminists hopped on
board and then their public support no
conflict was ever looking certain until
the feminist said we oppose this and
then H an hour we'll stop proclaiming
that this is for women's rights never
this is important why is this
necessarily important on a comparative
because that they're very best at their
very best they are slightly less likely
to go into these wars right all the
reasons pentalogy about how people
already proclaim that they're acting on
the rights of women in terms of a sham
Proclamation this doesn't occur and
they're worse than with their homes you
harm the perception of feminists at home
on our side the conflict still happens
as it will already except Caitlin Moran
in the Fawcett Society or I don't know
feminist organization in other countries
because that's kind of the point right
there are peripheral organizations with
very little international sway and you
get enormous harms to the feminist
movement which the we think needed to be
weighted in this debate

 

ちょこちょこ違うところはありますが、だいたい合っていると思います。

 

 

(2)活用方法

2つ活用方法を紹介したいと思います。

一つ目はシンプルに海外の大会動画やレクチャー動画をYoutubeで見る時の補助にする方法。

難しめの語彙やスピードでも結構拾ってくれるので、英語がそんなに得意じゃなくても効率的に音源を聴くことができます。

 

 

二つ目は自分のスピーチをYoutubeがどのくらい字幕にしてくれるかチェックするという使い方です。

自動字幕生成は発音が良ければスピードがいくら早くても生成してくれますが、日本人的な英語の発音には厳しいので、

これを利用すると自分のスピーチが日本人以外からどのように聞こえるのかを擬似的に学ぶことができます。

 

ためしに自分のある大会のスピーチの自動生成字幕を貼ってみます。

恥ずかしいので動画は貼りません。

 

greetings (→three thingsって言ってる)
Hamas cost revenues sto that invoice

chemo indignant on so dirty sentiment a
second Rita disposable breast operation(→corporationって言ってる)
sorry(→thirdlyって言ってる) de Verano sentiment really really
special on Impala meant a Odaiba
derivative a party(→firstlyって言ってる) in town necessity of
the mo+ of intonation Estrada daddy
Alberto provision Brockton a broccoli(→practicalって言ってる)
that we need a decimal we don't eat soda(→we don't think soって言ってる)
actually the most of the social women
fail because they create a fictional
settlement and require regular hottest
moments they're not that they require
their wall station that is the super
loop the initiation of a larger animal
movement it was a little is a predicate
animal rights society Globus conduct is
not only technology Morrison
pre-colonization(→principleって言ってる) travelers can track of

 

これはひどい

字幕を見ただけでthとか色々な英語的発音ができていないのがわかりますね。

限定公開でもたしか字幕つけてくれるはずなので、自分のスピーチなりラウンドなりをアップして、復習兼発音矯正みたいな使い方ができるんじゃないかと思います。