Ajit Narayanan: A word game to communicate in any language

譯者: Marssi Draw 審譯者: Willy Feng

I work with children with autism.

我服務有自閉症的孩子。

Specifically, I make technologies

更確切來說，我發明科技

to help them communicate.

幫助他們溝通。

Now, many of the problems that children

許多自閉症孩童面臨的問題

with autism face, they have a common source,

出自於同樣的因素，

and that source is that they find it difficult

那就是他們很難

to understand abstraction, symbolism.

了解抽象概念與象徵性的符號。

And because of this, they have a lot of difficulty with language.

因此，他們在面對語言時 會有很大的困難。

Let me tell you a little bit about why this is.

讓我告訴你一些原因。

You see that this is a picture of a bowl of soup.

你可以看到這張圖片是一碗湯。

All of us can see it. All of us understand this.

我們每個人都看得見，也都了解這是什麼。

These are two other pictures of soup,

這是另外兩張湯的圖片，

but you can see that these are more abstract

但是你會發現它們比較抽象，

These are not quite as concrete.

不太具體。

And when you get to language,

當你使用語言時，

you see that it becomes a word

會發現那個字詞

whose look, the way it looks and the way it sounds,

看起來、聽起來

has absolutely nothing to do with what it started with,

和它以什麼開頭

or what it represents, which is the bowl of soup.

或是和它代表的意義「那碗湯」完全無關。

So it's essentially a completely abstract,

因此，基本上那是一個完全抽象、

a completely arbitrary representation of something

存在真實世界中某種事物的

which is in the real world,

一種任意的表述，

and this is something that children with autism

自閉症的孩子在這方面

have an incredible amount of difficulty with.

有很大的困難。

Now that's why most of the people that work with children with autism --

那就是為什麼許多 協助自閉症孩童的人們

speech therapists, educators --

——語言治療師、教育人士——

what they do is, they try to help children with autism

他們協助自閉症孩童

communicate not with words, but with pictures.

不是用文字溝通，而是用圖片溝通。

So if a child with autism wanted to say,

因此如果有個自閉症孩童想說：「我想喝湯。」

"I want soup," that child would pick

這孩子會拿起

three different pictures, "I," "want," and "soup,"

三張不同的圖片「我」、「想喝」、「湯」，

and they would put these together,

然後把圖排在一起，

and then the therapist or the parent would

那麼治療師或家長就能理解

understand that this is what the kid wants to say.

這是孩子想說的話。

And this has been incredibly effective;

三四十年來

for the last 30, 40 years

這方法一直都很有效，

people have been doing this.

大家都這麼做。

In fact, a few years back,

事實上，幾年前

I developed an app for the iPad

我開發了一個 iPad 的應用程式，

which does exactly this. It's called Avaz,

名為「阿維思」(Avaz)，就是採用此法。

and the way it works is that kids select

操作方式是讓孩子選擇

different pictures.

不同的圖片，

These pictures are sequenced together to form sentences,

將圖片排列成句子，

and these sentences are spoken out.

然後這些句子會被唸出。

So Avaz is essentially converting pictures,

因此基本上「阿維思」會轉換圖片，

it's a translator, it converts pictures into speech.

它是翻譯機，能將圖片轉換成言語。

Now, this was very effective.

這很有用。

There are thousands of children using this,

有成千上萬的孩子使用它，

you know, all over the world,

遍及全世界，

and I started thinking about

於是我開始思考

what it does and what it doesn't do.

它做了什麼，又漏了什麼。

And I realized something interesting:

我發現某件很有趣的事：

Avaz helps children with autism learn words.

「阿維思」協助有自閉症的孩子學習文字。

What it doesn't help them do is to learn

但沒有教他們

word patterns.

文字模式。

Let me explain this in a little more detail.

讓我說明一些細節。

Take this sentence: "I want soup tonight."

以此句為例：「我今晚想喝湯。」

Now it's not just the words here that convey the meaning.

這不只是文字傳達了意義，

It's also the way in which these words are arranged,

這些文字排列的方式、

the way these words are modified and arranged.

這些文字修飾與排列的方式也有意義。

And that's why a sentence like "I want soup tonight"

那就是為什麼像是「我今晚想喝湯」這句話

is different from a sentence like

會完全不同於

"Soup want I tonight," which is completely meaningless.

「湯想喝我今晚」這樣無意義的句子。

So there is another hidden abstraction here

這裡有另一種隱藏的抽象概念，

which children with autism find a lot of difficulty coping with,

讓自閉症孩童難以處理，

and that's the fact that you can modify words

那就是你能透過修飾文字、

and you can arrange them to have

排列文字，

different meanings, to convey different ideas.

讓它有不同的意義，傳達不同的想法。

Now, this is what we call grammar.

我們稱之為文法。

And grammar is incredibly powerful,

而文法的力量十分強大，

because grammar is this one component of language

因為文法是語言的其中一項要素，

which takes this finite vocabulary that all of us have

讓我們使用所擁有的有限字彙

and allows us to convey an infinite amount of information,

傳達無限種資訊、

an infinite amount of ideas.

無限種想法。

It's the way in which you can put things together

這種方式能讓你把東西組合在一起

in order to convey anything you want to.

來傳達所有你想表達的事。

And so after I developed Avaz,

因此在我開發「阿維思」之後，

I worried for a very long time

有件事讓我擔心很久，

about how I could give grammar to children with autism.

那就是我要怎麼教自閉症孩童文法。

The solution came to me from a very interesting perspective.

解決方式來自一種非常有趣的觀點。

I happened to chance upon a child with autism

我巧遇自閉症的孩童

conversing with her mom,

和她的母親對話，

and this is what happened.

事情就這樣發生了。

Completely out of the blue, very spontaneously,

事發非常突然、不期而遇，

the child got up and said, "Eat."

那孩子站起來說：「吃。」

Now what was interesting was

有趣的是

the way in which the mom was trying to tease out

那位媽媽誘導小孩的方式，

the meaning of what the child wanted to say

她讓小孩透過回答她的問題

by talking to her in questions.

表達出想說的話。

So she asked, "Eat what? Do you want to eat ice cream?

因此她問：「吃什麼？」 「你想吃冰淇淋？」

You want to eat? Somebody else wants to eat?

「你想吃？」 「其他人想吃？」

You want to eat cream now? You want to eat ice cream in the evening?"

「你想現在吃冰淇淋？」 「你想晚上吃冰淇淋？」

And then it struck me that

我突然意識到

what the mother had done was something incredible.

那位母親做了一件非常棒的事。

She had been able to get that child to communicate

她已經能讓那個孩子

an idea to her without grammar.

不用文法就能傳達想法。

And it struck me that maybe this is what

我突然想到也許這就是

I was looking for.

我在找的方式。

Instead of arranging words in an order, in sequence,

與其透過按照規則、順序 將文字排列成句子，

as a sentence, you arrange them

不如將文字排列在這張圖中，

in this map, where they're all linked together

文字連結在一起的方式

not by placing them one after the other

不是透過將它們一個接一個排列，

but in questions, in question-answer pairs.

而是透過問題，多組問答題。

And so if you do this, then what you're conveying

因此如果你這麼做，那你傳達的

is not a sentence in English,

不是一個英文句子，

but what you're conveying is really a meaning,

你傳達的是一個意義，

the meaning of a sentence in English.

一個英文句子的意義。

Now, meaning is really the underbelly, in some sense, of language.

從某個層面來說， 意義在語言中屬於較深層的部分。

It's what comes after thought but before language.

意義出現在想法之後，但是在語言之前。

And the idea was that this particular representation

而此想法是這種特殊的表述

might convey meaning in its raw form.

可能是用它的根本樣貌來傳達意義。

So I was very excited by this, you know,

這件事讓我很興奮，

hopping around all over the place,

開心得手舞足蹈，

trying to figure out if I can convert

試著確認我是否能

all possible sentences that I hear into this.

將所有聽見的詞句轉換成這樣。

And I found that this is not enough.

我發現這還不夠。

Why is this not enough?

為什麼不夠呢？

This is not enough because if you wanted to convey

不夠是因為如果你想要傳達

something like negation,

否定的句子，

you want to say, "I don't want soup,"

比如說：「我不想喝湯。」

then you can't do that by asking a question.

那麼你就不能用問句完成。

You do that by changing the word "want."

你會改變「想」這個字。

Again, if you wanted to say,

同樣地，如果你想說：

"I wanted soup yesterday,"

「我昨天本來 想喝湯。」

you do that by converting the word "want" into "wanted."

你把「想」轉換成「本來想」。

It's a past tense.

那是過去式。

So this is a flourish which I added

因此我加了這個功能

to make the system complete.

讓系統更完善。

This is a map of words joined together

這是許多單字的連結圖，

as questions and answers,

以問句和答案組合而成，

and with these filters applied on top of them

有了這些篩選功能在上面，

in order to modify them to represent

就能做修改，呈現出

certain nuances.

較細微的差異。

Let me show you this with a different example.

讓我舉個不同的例子來說明。

Let's take this sentence:

以這個句子來說：

"I told the carpenter I could not pay him."

「我告訴了木工我不能付錢。」

It's a fairly complicated sentence.

這是個蠻複雜的句子。

The way that this particular system works,

這個特殊系統運作的方式是

you can start with any part of this sentence.

你可以從句子的任何一處開始。

I'm going to start with the word "tell."

我用「告訴」開頭來做說明。

So this is the word "tell."

這個字是「告訴」，

Now this happened in the past,

但這是以前發生的事，

so I'm going to make that "told."

所以我要說「告訴了」。

Now, what I'm going to do is,

現在我想做的是，

I'm going to ask questions.

我開始問問題。

So, who told? I told.

是誰「告訴」？ 是我。

I told whom? I told the carpenter.

我告訴了誰？ 我告訴了木工。

Now we start with a different part of the sentence.

現在我們從句子的另一處開始，

We start with the word "pay,"

以「付錢」開始，

and we add the ability filter to it to make it "can pay."

我們加上使役動詞，讓它變成「能付錢」，

Then we make it "can't pay,"

接著我們就能改成「不能付錢」，

and we can make it "couldn't pay"

接著就能更改時態，

by making it the past tense.

將它改為過去式。

So who couldn't pay? I couldn't pay.

那是誰不能付錢？ 我不能付錢。

Couldn't pay whom? I couldn't pay the carpenter.

不能付錢給誰？ 我不能付錢給木工。

And then you join these two together

接著你透過問這個問題

by asking this question:

把這兩個部分連在一起：

What did I tell the carpenter?

我告訴了木工什麼？

I told the carpenter I could not pay him.

我告訴了木工我不能付錢。

Now think about this. This is

想想看這個問題，

—(Applause)—

（掌聲）

this is a representation of this sentence

這是這個句子要表達的內容，

without language.

沒有語言。

And there are two or three interesting things about this.

這裡有兩到三件有趣的事。

First of all, I could have started anywhere.

首先，我能從任何一個單字開始，

I didn't have to start with the word "tell."

我不一定要從「告訴」開始。

I could have started anywhere in the sentence,

我能從句子的任何一部分開始，

and I could have made this entire thing.

還是能完成整件事。

The second thing is, if I wasn't an English speaker,

第二點是，如果我不是說英語的人，

if I was speaking in some other language,

如果我說的是別的語言，

this map would actually hold true in any language.

這個地圖真的在任何語言都管用。

So long as the questions are standardized,

只要這個問題符合標準，

the map is actually independent of language.

這個地圖就能獨立於語言使用。

So I call this FreeSpeech,

因此我稱它為「輕鬆講」 (FreeSpeech)，

and I was playing with this for many, many months.

我已經玩了好幾個月，

I was trying out so many different combinations of this.

並試著使用許多不同的組合。

And then I noticed something very interesting about FreeSpeech.

後來，我注意到「輕鬆講」有個有趣的部分。

I was trying to convert language,

我試著轉換語言，

convert sentences in English into sentences in FreeSpeech,

轉換英語句子和「輕鬆講」的句子，

and vice versa, and back and forth.

來回反覆不斷嘗試。

And I realized that this particular configuration,

我理解這種特殊的結構，

this particular way of representing language,

這種表現語言的特殊方式

it allowed me to actually create very concise rules

讓我能夠真正地建立很簡要的規則，

that go between FreeSpeech on one side

在「輕鬆講」

and English on the other.

以及英語之間的規則。

So I could actually write this set of rules

我確實能寫下這組規則，

that translates from this particular representation into English.

讓這個特殊的表述轉換成英語。

And so I developed this thing.

因此我發明了這項產品，

I developed this thing called the FreeSpeech Engine

稱為「輕鬆講引擎」，

which takes any FreeSpeech sentence as the input

能把任何「輕鬆講」的句子輸入，

and gives out perfectly grammatical English text.

然後產出有完美文法的英語。

And by putting these two pieces together,

透過組合

the representation and the engine,

表述與引擎，

I was able to create an app, a technology for children with autism,

我就能建立一個應用程式， 一個供自閉症孩童用的科技，

that not only gives them words

不只是提供他們文字，

but also gives them grammar.

也提供他們文法。

So I tried this out with kids with autism,

我在自閉症孩童身上測試，

and I found that there was an incredible amount of identification.

發現了很驚人的成效。

They were able to create sentences in FreeSpeech

他們用「輕鬆講」建立的句子

which were much more complicated but much more effective

複雜程度和效用都遠高於

than equivalent sentences in English,

用英語講同一句話，

and I started thinking about

我開始思考

why that might be the case.

為什麼會成功。

And I had an idea, and I want to talk to you about this idea next.

因此，接下來我想與大家分享一個想法。

In about 1997, about 15 years back,

大約在 1997 年時，大約 15 年前，

there were a group of scientists that were trying

有一群科學家嘗試

to understand how the brain processes language,

理解大腦處理語言的方式，

and they found something very interesting.

他們發現一件很有趣的事情。

They found that when you learn a language

就是當你學習一種語言，

as a child, as a two-year-old,

身為一個兩歲小孩，

you learn it with a certain part of your brain,

你用大腦的特定部位在學習；

and when you learn a language as an adult --

而當你身為一名成人

for example, if I wanted to learn Japanese right now —

──舉例來說，如果我現在想學日語──

a completely different part of my brain is used.

就會運用完全不同部位的大腦。

Now I don't know why that's the case,

我不了解為什麼會這樣，

but my guess is that that's because

但我猜是因為

when you learn a language as an adult,

成年時學習語言

you almost invariably learn it

幾乎無可避免會

through your native language, or through your first language.

透過你的母語、習慣語言來學習。

So what's interesting about FreeSpeech

「輕鬆講」有趣的是

is that when you create a sentence

當你建立一個句子，

or when you create language,

或是建立一種語言，

a child with autism creates language with FreeSpeech,

自閉症孩童用「輕鬆講」建立語言，

they're not using this support language,

他們不是用它來支援語言，

they're not using this bridge language.

他們不是用它來連結語言，

They're directly constructing the sentence.

他們是直接建立句子。

And so this gave me this idea.

這讓我有個想法。

Is it possible to use FreeSpeech

有可能讓「輕鬆講」

not for children with autism

教自閉症孩童語言之外，

but to teach language to people without disabilities?

也教非身障的孩童嗎？

And so I tried a number of experiments.

因此我嘗試許多實驗。

The first thing I did was I built a jigsaw puzzle

首先我設計了一個拼圖，

in which these questions and answers

這些問題和答案

are coded in the form of shapes,

都編碼成各種形狀，

in the form of colors,

各種顏色，

and you have people putting these together

操作人把這些放在一起，

and trying to understand how this works.

試著了解這是如何運作。

And I built an app out of it, a game out of it,

我設計了一個應用程式，以此為基礎的遊戲，

in which children can play with words

孩童可以玩文字遊戲，

and with a reinforcement,

並且有強化的功能，

a sound reinforcement of visual structures,

以聽覺強化視覺，

they're able to learn language.

他們就能學習語言。

And this, this has a lot of potential, a lot of promise,

這有很大的潛力和前景，

and the government of India recently

而最近印度政府

licensed this technology from us,

向我們取得這項科技的授權，

and they're going to try it out with millions of different children

他們打算讓上百萬名孩童嘗試，

trying to teach them English.

試著教他們英語。

And the dream, the hope, the vision, really,

而這個夢想、希望、願景

is that when they learn English this way,

即是當他們以此學習英語，

they learn it with the same proficiency

他們能夠表達流利，

as their mother tongue.

就像母語一樣。

All right, let's talk about something else.

接下來，我們來討論另一點。

Let's talk about speech.

談談說話。

This is speech.

這是說話。

So speech is the primary mode of communication

因此說話是溝通的基礎，

delivered between all of us.

在我們之間傳遞訊息。

Now what's interesting about speech is that

關於說話，有趣的是

speech is one-dimensional.

說話是單面的。

Why is it one-dimensional?

為什麼是單面的？

It's one-dimensional because it's sound.

因為說話是聲音，所以它是單面的。

It's also one-dimensional because

也因為

our mouths are built that way.

那是嘴巴的功能。

Our mouths are built to create one-dimensional sound.

嘴巴的功能即是創造單面的聲音。

But if you think about the brain,

但是如果你想想大腦，

the thoughts that we have in our heads

在我們頭腦裡的思想

are not one-dimensional.

並非一面向的。

I mean, we have these rich,

我的意思是，我們有這些豐富、

complicated, multi-dimensional ideas.

複雜和多面向的想法。

Now, it seems to me that language

對我來說，語言

is really the brain's invention

就是大腦的發明，

to convert this rich, multi-dimensional thought

一方面轉換這豐富、

on one hand

多面向的思想，

into speech on the other hand.

另一方面轉換成話語。

Now what's interesting is that

有趣的是

we do a lot of work in information nowadays,

現在我們以資訊做許多事，

and almost all of that is done in the language domain.

幾乎所有的事情都是在語言的領域中完成。

Take Google, for example.

以 Google 為例，

Google trawls all these countless billions of websites,

Google 網羅千百萬個網站，

all of which are in English, and when you want to use Google,

全都是英語網站， 而當你想要用 Google，

you go into Google search, and you type in English,

進入 Google 搜尋功能列，輸入英語，

and it matches the English with the English.

會出現符合你要的英語。

What if we could do this in FreeSpeech instead?

有沒有可能我們改用「輕鬆講」這樣做呢？

I have a suspicion that if we did this,

我推測如果我們這麼做，

we'd find that algorithms like searching,

我們會發現一些規則系統，像是搜尋、

like retrieval, all of these things,

像是擷取，所有的這些功能

are much simpler and also more effective,

都更簡單也更有效，

because they don't process the data structure of speech.

因為他們不是處理說話的資料結構。

Instead they're processing the data structure of thought.

相反地，他們處理思想的資料結構。

The data structure of thought.

思想的資料結構。

That's a provocative idea.

那是個令人興奮的概念。

But let's look at this in a little more detail.

讓我們多深入看一點細節。

So this is the FreeSpeech ecosystem.

這是「輕鬆講」的生態系統。

We have the Free Speech representation on one side,

我們一邊有「輕鬆講」的畫面，

and we have the FreeSpeech Engine, which generates English.

另一邊也有「輕鬆講」的引擎產生英語。

Now if you think about it,

請想像

FreeSpeech, I told you, is completely language-independent.

「輕鬆講」是完全獨立的語言。

It doesn't have any specific information in it

裡面沒有任何關於英語的

which is about English.

特定資訊。

So everything that this system knows about English

因此對這個系統來說，

英語都已在引擎中編碼。

is actually encoded into the engine.

That's a pretty interesting concept in itself.

這之中有個很有趣的概念。

You've encoded an entire human language

你已經將所有的人類語言編碼入

into a software program.

一套軟體中。

But if you look at what's inside the engine,

但是如果你看這個引擎的內部，

it's actually not very complicated.

會發現其實不複雜，

It's not very complicated code.

不是很複雜的編碼。

And what's more interesting is the fact that

更有趣的是，

the vast majority of the code in that engine

在那個引擎中大多數的編碼

is not really English-specific.

其實都不是只針對英語。

And that gives this interesting idea.

因此有了這個有趣的想法，

It might be very easy for us to actually

我們也許可以因此輕易地

create these engines in many, many different languages,

建立很多很多不同語言的引擎，

in Hindi, in French, in German, in Swahili.

印度語、法語、德語、斯瓦希里語。 （註：斯瓦希里語是非洲使用人數最多的語言之一）

And that gives another interesting idea.

這引起了另一個有趣的想法。

For example, supposing I was a writer,

舉例來說，假設我是作家，

say, for a newspaper or for a magazine.

在報社或雜誌社工作。

I could create content in one language, FreeSpeech,

我的文章可以用一種語言「輕鬆講」來寫，

and the person who's consuming that content,

然後有個人買了那則報導，

the person who's reading that particular information

閱讀資訊的那個人

could choose any engine,

可以選擇任何引擎，

and they could read it in their own mother tongue,

他們可以用自己的母語閱讀，

in their native language.

用他們當地的語言閱讀。

I mean, this is an incredibly attractive idea,

我的意思是，這是非常吸引人的想法，

especially for India.

尤其是在印度。

We have so many different languages.

我們有好多種語言。

There's a song about India, and there's a description

有首關於印度的歌，其中有一段描述

of the country as, it says,

將國家比喻為

(in Sanskrit).

（梵語）。

That means "ever-smiling speaker

意謂著「使用美好語言、

of beautiful languages."

永遠微笑的講者」。

Language is beautiful.

語言是美好的。

I think it's the most beautiful of human creations.

我認為語言是人類最美好的創造。

I think it's the loveliest thing that our brains have invented.

我認為語言是人腦發明最可愛的東西。

It entertains, it educates, it enlightens,

語言能娛樂、教育、啟發，

but what I like the most about language

但是我最愛的一點是

is that it empowers.

語言能賦予力量。

I want to leave you with this.

我想分享一件事。

This is a photograph of my collaborators,

這是我合作夥伴的照片，

my earliest collaborators

我最初的合作夥伴，

when I started working on language

當我開始研究語言、

and autism and various other things.

自閉症和各種不同的事。

The girl's name is Pavna,

這位女孩名為帕芙娜，

and that's her mother, Kalpana.

那是她的母親卡派納，

And Pavna's an entrepreneur,

帕芙娜是企業家，

but her story is much more remarkable than mine,

但是她的故事比我的更非凡，

because Pavna is about 23.

因為帕芙娜大概才 23 歲。

She has quadriplegic cerebral palsy,

她患有四肢型腦性麻庳，

so ever since she was born,

因此從她出生以來，

she could neither move nor talk.

她就不能動也不能說話。

And everything that she's accomplished so far,

迄今她所完成的所有事情，

finishing school, going to college,

完成學業、上大學、

starting a company,

開公司，

collaborating with me to develop Avaz,

和我合作開發「阿維思」，

all of these things she's done

她要做任何事情

with nothing more than moving her eyes.

都只能移動她的雙眼。

Daniel Webster said this:

丹尼爾．韋伯斯特說： （註：美國已故政治家）

He said, "If all of my possessions were taken

「如果要拿走我的一切，

from me with one exception,

只能留下一種，

I would choose to keep the power of communication,

我會選擇保留溝通的能力，

for with it, I would regain all the rest."

以此，我就能取回全部。」

And that's why, of all of these incredible applications of FreeSpeech,

那就是「輕鬆講」的所有美好功能中，

the one that's closest to my heart

最能貼近我心的一種

still remains the ability for this

還保留這項能力，

to empower children with disabilities

賦予身障孩童

to be able to communicate,

擁能溝通的能力，

the power of communication,

擁有溝通的力量，

to get back all the rest.

就能取回一切。

Thank you.

謝謝。

(Applause)

（掌聲）

Thank you. (Applause)

謝謝。（掌聲）

Thank you. Thank you. Thank you. (Applause)

謝謝。（掌聲）

Thank you. Thank you. Thank you. (Applause)

謝謝。（掌聲）