- Добро пожаловать на Лингвофорум.
Автор Ayndryl
- января 10, 2011, 20:49antbez,
Eddag-1240 and Elyam-2200 are not programs. They are the label that we assign to the output, once the system reaches a stable minimum. The system we are developing is called "Nodespaces", and it is coupled to a statistical mechanical program called "Vectorial".
What they do is implementing the Ising model, and all-time classic of statistical mechanical modelling. Originally, it was proposed to explain ferromagnetism, but the model is so powerful that it has over the years been adapted and applied to the study of a variety of phenomena, including language contact and language drift. "Vectorial" is a dynamical system, that is, a rule that given one point in a state space, it tells us where it is going to be a unit of time later. The basic mathematical model is thus a manifold M, a map f:M-->M and a function s:M-->R.
Now, M represents all the possible states of the system, while f represents the time evolution, a rule that determines what is the state one unit of time later. For many practical cases, after some time the system converges to a subset of M of low dimension, usually called the "attractor". The partition function is controlled by a so-called critical exponent (beta). The numbers 1240 and 2200 means that we ran the system for beta=2.2 and beta=1.240. Note that a small change in the critical exponent yields a totally different language for the same input.
Your question on how are they connected with language can now be answered: as we have records of, say, Old Welsh, Middle Welsh, and Modern Welsh, then we have a traceable history of how Welsh changed. This is our Welsh manifold M. The rules defining each state is given, roughly, by linguists (phonology rules, consonantal conservation, pallatization, etc.). Therefore, you can code the system to apply those rules to Modern Welsh and see what comes out. Of course, human brain is not a logical system: it is a highly complex and strange system. Nobody knows why a given change started operating in a given language, or why speakers gradually start modifying the language. Also, cultural and cognitive factors must be called for. This means we are facing a system that cannot be specified in an algorithmic way, neither using a rule-based approach. You need to use statistical mechanical maths.
Alexandra,
You pose the question of whether is it possible the transformation of Welsh into a language so different and unintelligible for today's speakers during a period so short as sixty-five years. How much time does a language take to change depends on many extra-linguistic factors. Isolation, the foundation effect, language contact, speakers' fragmentation in space, even religious, political, and social factors do all play a role. There is no definite answer to why some languages die, or to why they change the way they do.
But no, I'm sure 65 years is too a short time lapse for a radical change as the one you notice from Welsh to Weddag-2075. But the intriguing question is this: though they are structurally different, and lexically unrelated, what's the "Welshness" in Weddag-2075 as to make some people believe it is some sort of Welsh?
Demetrius,
I totally agree with the conclusion of the article, that is: "Russian, as well as any other language, despite it's obvious and unquestionable virtues, is by no means an ideal mirror for reflecting reality". In fact, there is a school of thought that holds that language "infects" reality.
You state that you you don't believe that feelings and words are directly related. Margaret Magnus, and the school of phonosemantics -aka sound symbolism-, has made interesting experiments with students to which she presents images of objects and ask them to label them with words. Some of them are impossible objects, or even non-existent, and the students are required to invent words for those objects. Surprisingly, there is a preference for using words with specific phonemes for certain objects. Her dissertation is available in the internet (it is titled Phonosemantics: What's in a word?).
Your statement related to the dependency of words on culture has been, and is, thoroughly researched by Anna Wierzbicka. You can find her books in the internet, too. My example on Russian was taken from her. Your question on whether there is any proof that the feelings of a Lakota and a Russian are different in any way has a positive answer: there are proofs. There are cognitive tests that consistently prove a speaker of a certain language perceives reality in a different way than a speaker of another language. I know this sounds somehow shocking, but it is that way.
I don't understand your last statement: "language doesn't neccessarily correspond to our feelings". Unless a speaker wishes to lie, we have to accept that anything that it is said it is also meant, I mean, you say what you mean, otherwise communication would be impossible. Your article makes the point: there are feelings for which we do not have words. There are emotions for which we have no words. But the results from Wierzbicka's research is even more radical: there are speakers from a certain language that do have emotions that speakers of another language do not. This is baffling, but true.
NB: if anyone here find it difficult to get the books, I will be pleased to post the links.
Eddag-1240 and Elyam-2200 are not programs. They are the label that we assign to the output, once the system reaches a stable minimum. The system we are developing is called "Nodespaces", and it is coupled to a statistical mechanical program called "Vectorial".
What they do is implementing the Ising model, and all-time classic of statistical mechanical modelling. Originally, it was proposed to explain ferromagnetism, but the model is so powerful that it has over the years been adapted and applied to the study of a variety of phenomena, including language contact and language drift. "Vectorial" is a dynamical system, that is, a rule that given one point in a state space, it tells us where it is going to be a unit of time later. The basic mathematical model is thus a manifold M, a map f:M-->M and a function s:M-->R.
Now, M represents all the possible states of the system, while f represents the time evolution, a rule that determines what is the state one unit of time later. For many practical cases, after some time the system converges to a subset of M of low dimension, usually called the "attractor". The partition function is controlled by a so-called critical exponent (beta). The numbers 1240 and 2200 means that we ran the system for beta=2.2 and beta=1.240. Note that a small change in the critical exponent yields a totally different language for the same input.
Your question on how are they connected with language can now be answered: as we have records of, say, Old Welsh, Middle Welsh, and Modern Welsh, then we have a traceable history of how Welsh changed. This is our Welsh manifold M. The rules defining each state is given, roughly, by linguists (phonology rules, consonantal conservation, pallatization, etc.). Therefore, you can code the system to apply those rules to Modern Welsh and see what comes out. Of course, human brain is not a logical system: it is a highly complex and strange system. Nobody knows why a given change started operating in a given language, or why speakers gradually start modifying the language. Also, cultural and cognitive factors must be called for. This means we are facing a system that cannot be specified in an algorithmic way, neither using a rule-based approach. You need to use statistical mechanical maths.
Alexandra,
You pose the question of whether is it possible the transformation of Welsh into a language so different and unintelligible for today's speakers during a period so short as sixty-five years. How much time does a language take to change depends on many extra-linguistic factors. Isolation, the foundation effect, language contact, speakers' fragmentation in space, even religious, political, and social factors do all play a role. There is no definite answer to why some languages die, or to why they change the way they do.
But no, I'm sure 65 years is too a short time lapse for a radical change as the one you notice from Welsh to Weddag-2075. But the intriguing question is this: though they are structurally different, and lexically unrelated, what's the "Welshness" in Weddag-2075 as to make some people believe it is some sort of Welsh?
Demetrius,
I totally agree with the conclusion of the article, that is: "Russian, as well as any other language, despite it's obvious and unquestionable virtues, is by no means an ideal mirror for reflecting reality". In fact, there is a school of thought that holds that language "infects" reality.
You state that you you don't believe that feelings and words are directly related. Margaret Magnus, and the school of phonosemantics -aka sound symbolism-, has made interesting experiments with students to which she presents images of objects and ask them to label them with words. Some of them are impossible objects, or even non-existent, and the students are required to invent words for those objects. Surprisingly, there is a preference for using words with specific phonemes for certain objects. Her dissertation is available in the internet (it is titled Phonosemantics: What's in a word?).
Your statement related to the dependency of words on culture has been, and is, thoroughly researched by Anna Wierzbicka. You can find her books in the internet, too. My example on Russian was taken from her. Your question on whether there is any proof that the feelings of a Lakota and a Russian are different in any way has a positive answer: there are proofs. There are cognitive tests that consistently prove a speaker of a certain language perceives reality in a different way than a speaker of another language. I know this sounds somehow shocking, but it is that way.
I don't understand your last statement: "language doesn't neccessarily correspond to our feelings". Unless a speaker wishes to lie, we have to accept that anything that it is said it is also meant, I mean, you say what you mean, otherwise communication would be impossible. Your article makes the point: there are feelings for which we do not have words. There are emotions for which we have no words. But the results from Wierzbicka's research is even more radical: there are speakers from a certain language that do have emotions that speakers of another language do not. This is baffling, but true.
NB: if anyone here find it difficult to get the books, I will be pleased to post the links.
Автор Demetrius
- января 10, 2011, 18:18Offtop
Цитата: Ayndryl от января 10, 2011, 09:04Really‽ I know nothing of quantum mechanics, but I have an impression that this area is a difficult to understand.
And all of them can easily understand quantum mechanics. This means humans are also complex adaptative systems.
Цитата: Ayndryl от января 10, 2011, 09:04I'm not quite sure I've understand your idea, but I believe that you don't take into consideration the inertness of the language. Languages do not change overnight. Surely, some part of the language really describes the environment we live in — but there are a great number of words that are passed down to us.
"Feeling Russian" in this context means this: Russian speakers have the words "duša" and "sud'ba", for which no translation exists in English. Translating them into "soul" and "fate" is just a rough approximation. Only a native speaker of Russian can fully grasp the meaning of "duša". Only a native speaker of Hebrew can grasp the full meaning of "memrah" or "ruach". Only a native speaker of Lakota can fully understand the word "wakantanga". And all of them can easily understand quantum mechanics. This means humans are also complex adaptative systems.
There are no "genuine" Russian (or any other language) words. There are genuine Russian (or some other) feelings. This is so because you shape your language to talk about reality and what you feel when facing that reality, but speakers do live in quite different areas on this planet and they experience quite different "microrealities", therefore it is but normal to see differences in how they perceive reality and, thus, in how they name those experiences. For a Lakota there is no "sud'ba" at all. He never experienced the need of it, thus no need to name it.
Sure, if you were born in Dakota in, say, 1783, and your language be Russian, you will need to invent a word for "wakantaga". But sure, too, your word "sud'ba" would then disappear, as it would describe something unexsistent within your Lakota world.
In fact, the lack of words for the feelings is a very common theme in poetry, regardless of the language (correct me if I'm wrong). And this idea is voiced not only in poetry. Below is an example taken from an article (I've tried to translate it; please excuse my English ^^).
Цитата: http://gramota.ru/biblio/magazines/nauka_i_zhizn/28_653
Indeed, in Russian one can find not only, so to speak, "useless words", i.e. words designating the same phenomenon in different ways. At the same time one can easily find the "obscure" parts of reality, that have to be filled, but there are no words for this. Here are some examples.
How can you name the feeling that a person has when they hear a fair criticism of their nearest and dearest? Of course it's протест (protest), гнев (gnev), возмущение (vozmuščenie), негодование (negodovanie). But if the criticism is grounded, the feelings are more complex and conflicting, but there is no words to describe them. (Смешанное чувство (smešannoe čuvstvo) is a too broad term.)
The person who gets on a public transport without a ticket is a заяц (zajac). But who is the person who bought a tickets for balcony and tries to get a seat in the stalls? It seems that this very common situation has no word for it too. We have a word завидовать (zavidovat'). It describes a situation when X has A, and Y, not having A, has negative feelings towards X. One can speak about the чёрная зависть (čërnaja zavist'), when the Y's negative feelings are so strong that (s)he may wish X the worst misfortunes. But there is also a белая зависть (belaja zavist'), when Y simply wants the thing X has, without having any negative feelings towards X. However, we can imagine a situation when X has A, and Y, not having A, is glad, because (s)he thinks such a situation is fair. "X раз за Y (X rad za Y)" doesn't accurately describe the situation, because it doesn't say anything about A that Y lacks. There is probably no way to describe the situation concisely and precisely.
I'd also like to remind you of the ambiguity that is not always desirable when we use words могу (mogu) or должен (dolžen). It's often unclear whether the moral or physical aspects are meant. And about the ambiguity of the duration of an action that is associted with many non-finite verbs. "Он ленится (on lenits'a)" — morally? Physically? Currently? During a some period? Always? "Он танцует (on tancujet)" — currently? Often?
To cut the long story short, Russian, as well as any other language, despite it's obvious and unquestionable virtues, is by no means an ideal mirror for reflecting reality. Our mirror has not only some "useless" parts, but also some obscure places.
I believe that feelings and words are not directly related. The words are depend on our culture. And if the knowledge of these words has been passed down to us, why can't the people of different nations gain this knowledge it too?
But the feelings... Is there any proof the feelings of a Lakota and a Russian are different in any way? Linguistic data about it is clearly not enough, because the language doesn't neccessarily correspond to our feelings.
I'm quite sure that if Lakota happened to live in Moskow, he would still have situations when the word "wakantaga" is appropriate.
In general, I find your experiment very interesting. Would you please write here when you have published it?
Автор Alexandra A
- января 10, 2011, 16:45Цитата: Ayndryl от января 10, 2011, 09:04A ydyw'n bosibl y trawsnewidiad o'r Gymraeg i'r iaith mor wahanol ac annealladwy â heddiw (a welais ar eich safle chwi) o fewn cyfnod mor fyr â phump a thrigain o flynyddoedd?
The one you thought to be Welsh is called Weddag-2075: we fed the system with Welsh, supressed the Normand contribution, added some Goidelic patterns, and left the system evolve
Is it possible, the transformation of Welsh into a language so different and unintelligible for today's speakers (as I saw on your site) during a period so short as sixty-five years?
Автор Квас
- января 10, 2011, 15:51Ayndryl, thanks for your detailed answer. 
Now your point is quite clear. Your investigation looks fascinating!
Now your point is quite clear. Your investigation looks fascinating!Автор antbez
- января 10, 2011, 15:45Цитировать
However, it is "attractors" and "homology" what we are talking here. As you know, it is imposible to find the original seed state S(0) of a Markov process given the final state S(f) by virtue of the stochastic nature of the process. But it might be of interest to you that, regardless of the initial state (that is, language) you arrive to quite stable (and beautiful) consistent configurations (like Eddag-1240 or Elyam-2200).
Thank You for the answer!
I haven't heard about Eddag-1240 or Elyam-2200. What are these programs? How are they connected with language?
Автор Ayndryl
- января 10, 2011, 09:04Kbas and antbez:
"Feeling Russian" in this context means this: Russian speakers have the words "dusa" and "sud'ba", for which no translation exists in English. Translating them into "soul" and "fate" is just a rough approximation. Only a native speaker of Russian can fully grasp the meaning of "dusa". Only a native speaker of Hebrew can grasp the full meaning of "memrah" or "ruach". Only a native speaker of Lakota can fully understand the word "wakantanga". And all of them can easily understand quantum mechanics. This means humans are also complex adaptative systems.
There are no "genuine" Russian (or any other language) words. There are genuine Russian (or some other) feelings. This is so because you shape your language to talk about reality and what you feel when facing that reality, but speakers do live in quite different areas on this planet and they experience quite different "microrealities", therefore it is but normal to see differences in how they perceive reality and, thus, in how they name those experiences. For a Lakota there is no "sud'ba" at all. He never experienced the need of it, thus no need to name it.
Sure, if you were born in Dakota in, say, 1783, and your language be Russian, you will need to invent a word for "wakantaga". But sure, too, your word "sud'ba" would then disappear, as it would describe something unexsistent within your Lakota world.
We are not interested in proto-languages. We are interested in "proto-humans as they spoke" in order to grasp "post-humans as they will speak".
It is extremely easy to "invent" a language from scratch. However, as time passes, that language will evolve. And it happens it evolves in a quite specific and predictable way. The final state is always the same for any given language. This evolution is proto-language indpendent. We are more interested in tomorrow's language.
The length of the sentences, the speed at which you can talk a language, the phonology, the order of the sentence, all this is limited and constrained by your cognitive structure and your neurological setup. All languages will finally converge in time into one and the same language, provided reality remains the same.
But speakers, their languages, and reality itself are dynamical systems, always changing. These are non-linearties that modify the language and your cognitive structure in radical ways. The net result is this: new languages emerge, others they die.
So finally, we are investigating the state equation of language, assessing whether it is a Markov process, an Ising-spin process, a fractal process, or whatever. This is key for us in order to make both backward and forward language analysis.
The languages you happened to spot in our blog are final or intermediate states for a given scenario we are currently working on. The one you thought to be Welsh is called Weddag-2075: we fed the system with Welsh, supressed the Normand contribution, added some Goidelic patterns, and left the system evolve assigning specific value to the Beta exponent of an Ising system to see what happens. Beta = 2.075 in this particular scenario.
The language called Alashi-1330 is a semitic one with bi-radical nouns and IE SVO typology with a Lyapunov exponent of 1.33. And so on.
antbez,
Actually you can establish some isomorphism between language L1 and L2, though in general this does not work, as you correctly stated. However, it is "attractors" and "homology" what we are talking here. As you know, it is imposible to find the original seed state S(0) of a Markov process given the final state S(f) by virtue of the stochastic nature of the process. But it might be of interest to you that, regardless of the initial state (that is, language) you arrive to quite stable (and beautiful) consistent configurations (like Eddag-1240 or Elyam-2200).
Finally, the configurations you arrive at are all languages. You can use them as you use your native ones, you can use them to describe reality or write poetry. What you see in our blog are the entries of the users where they announce the translation of the relevant books into a given generated language.
Again, we are not interested in "conlangs". We are interested in fractal linguistics.
"Feeling Russian" in this context means this: Russian speakers have the words "dusa" and "sud'ba", for which no translation exists in English. Translating them into "soul" and "fate" is just a rough approximation. Only a native speaker of Russian can fully grasp the meaning of "dusa". Only a native speaker of Hebrew can grasp the full meaning of "memrah" or "ruach". Only a native speaker of Lakota can fully understand the word "wakantanga". And all of them can easily understand quantum mechanics. This means humans are also complex adaptative systems.
There are no "genuine" Russian (or any other language) words. There are genuine Russian (or some other) feelings. This is so because you shape your language to talk about reality and what you feel when facing that reality, but speakers do live in quite different areas on this planet and they experience quite different "microrealities", therefore it is but normal to see differences in how they perceive reality and, thus, in how they name those experiences. For a Lakota there is no "sud'ba" at all. He never experienced the need of it, thus no need to name it.
Sure, if you were born in Dakota in, say, 1783, and your language be Russian, you will need to invent a word for "wakantaga". But sure, too, your word "sud'ba" would then disappear, as it would describe something unexsistent within your Lakota world.
We are not interested in proto-languages. We are interested in "proto-humans as they spoke" in order to grasp "post-humans as they will speak".
It is extremely easy to "invent" a language from scratch. However, as time passes, that language will evolve. And it happens it evolves in a quite specific and predictable way. The final state is always the same for any given language. This evolution is proto-language indpendent. We are more interested in tomorrow's language.
The length of the sentences, the speed at which you can talk a language, the phonology, the order of the sentence, all this is limited and constrained by your cognitive structure and your neurological setup. All languages will finally converge in time into one and the same language, provided reality remains the same.
But speakers, their languages, and reality itself are dynamical systems, always changing. These are non-linearties that modify the language and your cognitive structure in radical ways. The net result is this: new languages emerge, others they die.
So finally, we are investigating the state equation of language, assessing whether it is a Markov process, an Ising-spin process, a fractal process, or whatever. This is key for us in order to make both backward and forward language analysis.
The languages you happened to spot in our blog are final or intermediate states for a given scenario we are currently working on. The one you thought to be Welsh is called Weddag-2075: we fed the system with Welsh, supressed the Normand contribution, added some Goidelic patterns, and left the system evolve assigning specific value to the Beta exponent of an Ising system to see what happens. Beta = 2.075 in this particular scenario.
The language called Alashi-1330 is a semitic one with bi-radical nouns and IE SVO typology with a Lyapunov exponent of 1.33. And so on.
antbez,
Actually you can establish some isomorphism between language L1 and L2, though in general this does not work, as you correctly stated. However, it is "attractors" and "homology" what we are talking here. As you know, it is imposible to find the original seed state S(0) of a Markov process given the final state S(f) by virtue of the stochastic nature of the process. But it might be of interest to you that, regardless of the initial state (that is, language) you arrive to quite stable (and beautiful) consistent configurations (like Eddag-1240 or Elyam-2200).
Finally, the configurations you arrive at are all languages. You can use them as you use your native ones, you can use them to describe reality or write poetry. What you see in our blog are the entries of the users where they announce the translation of the relevant books into a given generated language.
Again, we are not interested in "conlangs". We are interested in fractal linguistics.
Автор Bhudh
- января 10, 2011, 03:19Вручную, наверное.
Автор Demetrius
- января 10, 2011, 02:46Ух ты как интересно. 
Цитата: Ayndryl от декабря 31, 2010, 09:08Если честно, я не понял, как они собрались моделировать изменение семантики.
- to analyze syntactic and morphological drift for a given set of
languages, and to explore whether such a drift produces a semantic drift
correlate.
Автор Квас
- января 7, 2011, 19:50Ayndryl, your experiment is fairly interesting. And in case you happen to drop in once again, I'd like to ask you a couple of question.
I don't understand it either. Neither I understand what you mean by '"feel" like a Russian'. How can one tell if I feel like a Russian or not? What does such a feeling affect? That is are there any tests that allow to make out whether a given individual feels like a Russian/Lakota/etc?
You mean the Proto-Slavic language, don't you?
Цитата: Ayndryl от декабря 31, 2010, 09:08
Sure you can get a total mastering of the Russian language. But
you will never "feel" like a Russian. So the question arises: what do we
mean by being Russian or Lakota?
I don't understand it either. Neither I understand what you mean by '"feel" like a Russian'. How can one tell if I feel like a Russian or not? What does such a feeling affect? That is are there any tests that allow to make out whether a given individual feels like a Russian/Lakota/etc?
Цитата: Ayndryl от декабря 31, 2010, 09:08
And if there was just one
protolanguage, what made a given speaker to start feeling like a
Russian? The landscape? The environment? A genetic mutation? A specific
neurological arrangement?
You mean the Proto-Slavic language, don't you?
Автор antbez
- января 3, 2011, 19:45Я прекрасно понимаю, о чём идёт речь, но если вы- человек серьёзный, то должны понять, что автоматически "заменить" не получится. Это- не математика, где мы можем построить чёткие изоморфизмы (и то не всегда). То есть возможна некоторая схема, по которой можно что-то восстановить и в итоге догадаться об общем смысле рассматриваемой фразы. Но не более того... К тому и была моя реплика, что не стоит переусердствовать!