Question
Could we translate Speaking language to Conceptual language?
We all know when we hear any statement in my Native language our brain converts this statement to Conceptual language that constructed in our brain. Could we have an artificial translator that performs this conversion from any language? With this translator would we be able understand any Foreign language just like Native language?
All Answers (26)
-
Sounds like a fascinating idea. Maybe with the help of Brain-Computer Interfacing (BCI)? http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001251 -
Well, as far as i know, no artificial translation could ever behave like human translators simply due to the lack of cognition. perhaps som day... -
Actually, I would probably disagree with your starting assumption (that there is some "conceptual language" in our brain). The idea that humans share some common "language of thought" is not a new one, but there is little cog-sci evidence to support it, a fair amount of evidence to the contrary, and it does not fit well with the basic model of the brain as a massively 'parallel', spreading-activation network.
Brains appear to latch on to the patterns of language just the way they latch on to the patterns of human behavior (sometimes as near rule-governed sequences, mostly just weighted-probabilities). And each brain develops its own weighted-associations model of public linguistic structures, just as it has its own such model of human behaviors. We may come pre-wired with sensory discriminations (phoneme types and color distinctions, maybe even facial geometry, as firmware), but we certainly aren't born with anything like the 'concepts' necessary for a language capable of describing 'democracy', 'alarm clocks' or 'cheese-whiz'. We must build these concepts from experience, and as each of us has a different set of experiences, consequently we will each have nuanced differences in our concepts.
Conclusion: I think your suggestion is an interesting non-starter, but great question. -
In addition, there is evidence that the human brain uses multiple channels of language processing (see Giora on graded salience) and multiple conceptual contexts for interpretation / generation (see Kecskess on duelling contexts). -
by context-free Grammar for pronouncing sentences. G=(st,snt,P,S) where st=terminal symbosl set;snt=nonterminal symbols set;P=production rules and S=starting symbol.Result: language understandable -
Firstname Lastname: I agree with you that "We must build these concepts from experience, and as each of us has a different set of experiences, consequently we will each have nuanced differences in our concepts", but I think we must construct an adaptive translator, so it can learn conceptual language of any person Separately -
my own take on this is the following:
we are all prewired the same way. all material that goes into linguistic meanings must be common to all human beings. which is why my own belief is that there is no concept in any human lg that i or any human can't understand.
which is why there is pretty littlle that can't really be translated across languages. that part that hasn't found expresson in one lg is supressed iin those speakers. Bringing this to conscious awareness is what is meant by saying that we build concepts thru' experience.
giridhar -
Giridhar P.P : do you mean that we understand linguistic meaning of any word in the same way, but understand the concept of word in the sentence based on our experiences?
-
Different programming language can be compiled into the same IR,then the IR can be translated into the other language. Is your inspiration for the artificial translator comes from here?
-
Liu Yanzhao: As a computer engineer I knew this fact, but my Idea didn't come from here :)
-
Check out the language lojban. It is context-free, a spoken and written language.
(from Wikipedia)
- It has a grammar that is based on predicate logic, designed to express complex logical constructs precisely.
- It has no irregularities or ambiguities in spelling and grammar (although word derivation relies on arbitrary variant -forms). This gives rise to high intelligibility for computer parsing.
- It is designed to be as culturally neutral as possible.
- It allows highly systematic learning and use, compared to most natural languages.
- It possesses an intricate system of indicators which effectively communicate contextual attitudes or emotions.
- It does not have simplicity as a design criterion. -
Joachim Pimiskern: can you explain more? In this way we couldn't communicate with each other at all!
-
Journey from Language( any language) to Conceptual model( symbolic model: tangible and intangible ) is almost similar as we travel around the universe. But conversion from conceptual to speakable language is not that easy, the fundamental reason is the placement of symbols in different sequence(grammer + context ) is hard enough to solve 100% accurately. Be ware Natural language processing is hot research topic now a days, might be in future we come up with excitings. -
This image shows the difference between natural intelligence and artificial intelligence.
We have to understand more about how brain works in order to improve Natural language processing (NLP) -
Good.
this has been edifying,
My point is that we convert percepts to mental concepts before linguisticizing them.
If we are talking of natural lg, it is not right to say we can assign any meaning to any symbol. then ONLY you will understand it. LINGUISTIC STR IS NOT SOCIALLY DRIVEN BUT THE LEXICON IS.
To Younis Elhaddad
yours is agreat point about AI and NI.
How are the possibilities of a one yr engagement as professor of Engish in the uni of Bengazhi? i recently retired as a prof of linguistics in a premier national institute of indian languages in India. i have a reasonably compelling CV!
Giridhar -
To answer the question, should take into account that: a) There is a natural language and a conventional language, which have no relation to each other and b) there is no conceptual language.
Point b) is based on what is usually taken as thought, which is where the concepts are assumed, refers only to logical thinking. Logical thinking is that obey the laws of traditional logic and has nothing to do with the mental process of thinking.
Point a) requires greater detail. If it is of some interest, can be extended. -
When automated, you're describing the task of "machine translation" -- mapping the semantics of one natural language (NL) to another. One approach (among many) of an intermediate representation (IR) for natural language modeling and translation is "lexical semantics" (LS), <http://en.wikipedia.org/wiki/Lexical_semantics>;. The LS wiki page will give you some idea of what work has been done in this area and how such a NL IR model could work. More recent work on this seems to employ a lot of probability and statistics -- recognizing speech patterns and word associations rather than modeling deeper semantics. -
Randolph Crawford
Do you mean to use the same techniques of computer compilers and interpreters? -
Not really. Compiler/interpreter front ends are 98% based on context-free grammars (with a little supplemental annotation), making them amenable to implementation using traditional compiler tools like Lex and YACC. Programming languages are also extremely tiny, allowing explicit context-free grammar representations like Backus Naur Form to encompass all possible expressions. However natural languages are enormously larger and varied and irregular, and so cannot map neatly to a theoretical computational model nor a manageable, consistent set of rules. Most advances in machine translation or information extraction in the past decade seem to employ more statistical / probabilistic methods like n-grams, markov models, and part-of-speech tagging in an attempt to improve the signal-to-noise ratio during parsing and the extraction and mapping of semantics between languages. The rise of practical systems like Apple's Siri and IBM's Watson show that NL techniques have advanced significantly toward serving everyday tasks, demonstrating that these approaches are succeeding where more traditional and formal NLP models did not.
-
Thank you Sir. it is clear now
-
our brain converts this statement to Conceptual language that constructed in our brain. Could we have an artificial translator that performs this conversion from any language? Yes, I´m working on it
With this translator would we be able understand any Foreign language just like Native language? Yes, I¡m working on it, but takes times. What I saw till now, it is that we do not need the compiler, we need to create something new with the computer, and make it to understand the language based at the very beginning by itself, after this, you will have to create an engine to recognize pattern and combination of the words, orders, grammar criteria, and after you finished it, you have to make you whole brain system machine to have input and output to learn based on it. The key point it takes time, in fact google is working on it too, and many others too. Every one is using their own criteria and theory to make it works, my case this is my third year on it. Keep trying, u can use C++ as a tool to start by your own. -
The use of the word 'language' to describe systems for generating computer programs can be a little misleading. A computer 'programme' as presented to a computer is simply a list of machine instructions (for example set bits in register 2 to ones) that are executed sequentially by the machine. This means that the idea of the macihine 'understanding' the progamme has no obvious meaining in the terms that apply to people understanding a language. The only people who can understand a computer programme are other people. The brain is not obviously a 'machine' comparable with a computer in this sense and the analogy is not helpful in my opinion. In writing a computer programme we are going in the direction from human concepts - to instructions to a machine - to some physical result such as generating marks on a piece of paper or turning lights on and off.
-
I AGREE TOTALLY WITH GEOFFREY STEPHENSON'S REMARKS
A s for the conceptual lg coresponding to what we speak, there is no evidence for saying this. We speak of two interfaces that every linguistic expression must meet the condiitions of. viz the sensory-motor interface and the conceptual-intentional interface. For the conceptual lg to be true there needs to be a lg of thought that is distinct from what CI entails.
giridhar -
language is referential in its nature, question is do we have mathematical framework which can adequately model referential characteristic of "a" composition which represent a spoken word, a phrase or a paragraph?
I've tried to address that subject in my paper "The Harmonic Theory" let me know what you think
Popular Answers
We have to understand more about how brain works in order to improve Natural language processing (NLP)