What would it take to make a “true” universal translator?

[Maximum7]

Besides telepathy, what could I use to explain the properties of a device that is a TRUE universal translator, a handheld device that can decipher any language it encounters? I don’t know much about linguistics but I was told that even in a sci-fi setting, this is impossible. Jack Buchanan even said it would be harder to build the Enterprise than a universal translator.

 

[shjacks45]

Omniscience. To translate by definition requires sound(/light) correlation with ideas which requires either training or knowledge provided in advance (e.g. dictionary). Star Trek made the assumption that all humanoids had similar brain structure that language would follow. "All Earth languages evolved from Sanscrit" so someone knowing Sanscrit can deduce human language? Obviously not. And a few dozen languages derived from seemingly random roots. Then heiroglyphics/chinese which are right brain not left brain processes, with different assumptions of reality. Or the Filipino language with NO concept of time or tense.

 

[Vanadium 50]

A device that given a word or two can deduce all the other words and full grammar? Magic.

 

[jim mcnamara]

Noam Chomsky tried to solve the problem by postulating the existence of a Universal grammar.

Universal grammar (UG), in modern linguistics, is the theory of the genetic component of the language faculty, usually credited to Noam Chomsky. The basic postulate of UG is that a certain set of structural rules are innate to humans, independent of sensory experience.

-- Universal grammar - Wikipedia
https://en.wikipedia.org/wiki/Universal_grammar

As an aside, there was a NOVA program in the late nineties about the human language facility.
Two interesting points:
1. Native Japanese speakers use different brain structures for music and language than do people who learn Japanese in a multilingual environment. Or later in life as a second language.

2. Phonemes (basic sounds used to make words) apprehension may require hearing that phoneme in speech before the age of 12 months. After that learning window, new base phoneme acquisition degrades with time. This is, in part, the basis for foreign accents - second language speakers with problems with phonemes they did not encounter as kids.
Example:
There was a Native American language in Washington State (IIRC) that had a few phonemes no Indo-European language speaking brain could hear at all. That language no longer has any speakers. It was part of the the Salishan language family, a primary language group like Indo-European. So it had a lot of uniqueness.

 

[jim mcnamara]

PS: https://en.wikipedia.org/wiki/Salishan_languages
Read the article and note the use of bizarre English attempts at characters, like punctuation "?", in the names of these languages. Trying to make these odd phonemes have something written.

Also note that a common trait is tenselessness. Not a joke.

 

[berkeman]

PS: https://en.wikipedia.org/wiki/Salishan_languages
Read the article and note the use of bizarre English attempts at characters, like punctuation "?", in the names of these languages. Trying to make these odd phonemes have something written.

Also note that a common trait is tenselessness. Not a joke.

The Salishan (also Salish) languages are a group of languages of the Pacific Northwest in North America (the Canadian province of British Columbia and the American states of Washington, Oregon, Idaho and Montana).[2] They are characterised by agglutinativity and syllabic consonants. For instance the Nuxalk word clhp’xwlhtlhplhhskwts’ (IPA: [xɬpʼχʷɬtʰɬpʰɬːskʷʰt͡sʼ]), meaning "he had had [in his possession] a bunchberry plant", has thirteen obstruent consonants in a row with no phonetic or phonemic vowels.

 

[jim mcnamara]

I guess the point I failed to make: if you want to create a universal translator you would have to be able to create and perceive all ~126 human phonemes. Infants babble sounds are thought to have those phonemes in them. The 'neural wiring' for some phonemes gets lost as the kid fails to hear them in the language he/she hears and strengthened for the ones that are present in the languages the kid hears.

AFAIK, no one language uses all of them. Plus, the human race likely has lost and gained some along the way the past 100,000 years or so.

BTW the number 126 above was sort of a guess when I knew something about the literature long ago. Use with caution.

https://www.nidcd.nih.gov/health/speech-and-language

 

[jim mcnamara]

I was wrong - there are differing definitions of 'phoneme', so a fixed number of them does not work in general.
https://en.wikipedia.org/wiki/Phoneme