0:00:06.677,0:00:11.306
How is it that so many [br]intergalactic species in movies and TV

0:00:11.306,0:00:14.483
just happen to speak perfect English?

0:00:14.483,0:00:17.886
The short answer is that no one[br]wants to watch a starship crew

0:00:17.886,0:00:21.774
spend years compiling an alien dictionary.

0:00:21.774,0:00:23.392
But to keep things consistent,

0:00:23.392,0:00:26.789
the creators of Star Trek[br]and other science-fiction worlds

0:00:26.789,0:00:30.514
have introduced the concept[br]of a universal translator,

0:00:30.514,0:00:35.012
a portable device that can instantly[br]translate between any languages.

0:00:35.012,0:00:38.539
So is a universal translator [br]possible in real life?

0:00:38.539,0:00:42.137
We already have many programs[br]that claim to do just that,

0:00:42.137,0:00:45.954
taking a word, sentence, [br]or entire book in one language

0:00:45.954,0:00:49.004
and translating it into almost any other,

0:00:49.004,0:00:52.337
whether it's modern English[br]or Ancient Sanskrit.

0:00:52.337,0:00:55.913
And if translation were just a matter[br]of looking up words in a dictionary,

0:00:55.913,0:00:59.825
these programs would run circles[br]around humans.

0:00:59.825,0:01:03.299
The reality, however, [br]is a bit more complicated.

0:01:03.299,0:01:07.349
A rule-based translation program[br]uses a lexical database,

0:01:07.349,0:01:10.302
which includes all the words [br]you'd find in a dictionary

0:01:10.302,0:01:13.283
and all grammatical forms they can take,

0:01:13.283,0:01:18.925
and set of rules to recognize the basic[br]linguistic elements in the input language.

0:01:18.925,0:01:22.396
For a seemingly simple sentence like,[br]"The children eat the muffins,"

0:01:22.396,0:01:27.050
the program first parses its syntax,[br]or grammatical structure,

0:01:27.050,0:01:29.587
by identifying the children [br]as the subject,

0:01:29.587,0:01:32.317
and the rest of the sentence [br]as the predicate

0:01:32.317,0:01:34.368
consisting of a verb "eat,"

0:01:34.368,0:01:37.422
and a direct object "the muffins."

0:01:37.422,0:01:40.249
It then needs to recognize[br]English morphology,

0:01:40.249,0:01:44.681
or how the language can be broken down[br]into its smallest meaningful units,

0:01:44.681,0:01:46.124
such as the word muffin

0:01:46.124,0:01:49.755
and the suffix "s," [br]used to indicate plural.

0:01:49.755,0:01:52.449
Finally, it needs to understand [br]the semantics,

0:01:52.449,0:01:56.178
what the different parts of the sentence[br]actually mean.

0:01:56.178,0:01:58.074
To translate this sentence properly,

0:01:58.074,0:02:01.982
the program would refer to a different set[br]of vocabulary and rules

0:02:01.982,0:02:05.166
for each element of the target language.

0:02:05.166,0:02:07.020
But this is where it gets tricky.

0:02:07.020,0:02:11.820
The syntax of some languages[br]allows words to be arranged in any order,

0:02:11.820,0:02:16.954
while in others, doing so could make[br]the muffin eat the child.

0:02:16.954,0:02:19.647
Morphology can also pose a problem.

0:02:19.647,0:02:23.243
Slovene distinguishes between[br]two children and three or more

0:02:23.243,0:02:27.097
using a dual suffix absent [br]in many other languages,

0:02:27.097,0:02:30.532
while Russian's lack of definite articles[br]might leave you wondering

0:02:30.532,0:02:33.575
whether the children are eating [br]some particular muffins,

0:02:33.575,0:02:36.719
or just eat muffins in general.

0:02:36.719,0:02:39.708
Finally, even when the semantics[br]are technically correct,

0:02:39.708,0:02:42.757
the program might miss their finer points,

0:02:42.757,0:02:45.809
such as whether the children [br]"mangiano" the muffins,

0:02:45.809,0:02:47.794
or "divorano" them.

0:02:47.794,0:02:51.558
Another method is [br]statistical machine translation,

0:02:51.558,0:02:55.762
which analyzes a database [br]of books, articles, and documents

0:02:55.762,0:02:59.488
that have already [br]been translated by humans.

0:02:59.488,0:03:02.959
By finding matches between source[br]and translated text

0:03:02.959,0:03:05.393
that are unlikely to occur by chance,

0:03:05.393,0:03:09.345
the program can identify corresponding[br]phrases and patterns,

0:03:09.345,0:03:12.429
and use them for future translations.

0:03:12.429,0:03:14.969
However, the quality [br]of this type of translation

0:03:14.969,0:03:17.690
depends on the size [br]of the initial database

0:03:17.690,0:03:21.357
and the availability of samples [br]for certain languages

0:03:21.357,0:03:23.383
or styles of writing.

0:03:23.383,0:03:27.140
The difficulty that computers have[br]with the exceptions, irregularities

0:03:27.140,0:03:30.994
and shades of meaning[br]that seem to come instinctively to humans

0:03:30.994,0:03:35.045
has led some researchers to believe[br]that our understanding of language

0:03:35.045,0:03:39.251
is a unique product [br]of our biological brain structure.

0:03:39.251,0:03:43.101
In fact, one of the most famous[br]fictional universal translators,

0:03:43.101,0:03:46.439
the Babel fish from [br]"The Hitchhiker's Guide to the Galaxy",

0:03:46.439,0:03:49.726
is not a machine at all[br]but a small creature

0:03:49.726,0:03:54.210
that translates the brain waves[br]and nerve signals of sentient species

0:03:54.210,0:03:57.005
through a form of telepathy.

0:03:57.005,0:03:59.726
For now, learning a language[br]the old fashioned way

0:03:59.726,0:04:05.106
will still give you better results than[br]any currently available computer program.

0:04:05.106,0:04:06.749
But this is no easy task,

0:04:06.749,0:04:09.014
and the sheer number [br]of languages in the world,

0:04:09.014,0:04:12.989
as well as the increasing interaction[br]between the people who speak them,

0:04:12.989,0:04:18.004
will only continue to spur greater[br]advances in automatic translation.

0:04:18.004,0:04:21.409
Perhaps by the time we encounter[br]intergalactic life forms,

0:04:21.409,0:04:24.660
we'll be able to communicate with them[br]through a tiny gizmo,

0:04:24.660,0:04:29.026
or we might have to start compiling[br]that dictionary, after all.