Globobafflegab
« previous post | next post »
There's a "Quorum" (in the sense of a discussion among a select group) up at Freakonomics on the topic "What Will Globalization Do To Languages". I'm one of the participants. I tried, perhaps too hard, to find something non-obvious to say on this topic — maybe I just should have sounded the tocsin for preservation of endangered languages.
john riemann soong said,
May 29, 2008 @ 12:04 am
"The big loser? Grammar."
Christian Rolling seemed to be going good up until he said this.
Dick Margulis said,
May 29, 2008 @ 6:01 am
The classic line from the comments was this from "Jeff":
"When computers get fast enough, real-time translation will occur (where I can call up someone in China and they’ll hear me in their local dialect). This kind of thing is probably much closer than we realize."
Okay, I suppose we're closer than we were with Eniac, when the wish among linguists was for a computer with "a thousand words of memory so we could finally do machine translation right," as my freshman linguistics professor told the story back in 1964. Heck. Give it another 44 years and maybe we'll be there.
marie-lucie said,
May 29, 2008 @ 12:02 pm
Instant, real-time translation of speech sounds like a pipe dream of monolinguals. It is difficult enough to achieve accurate translation of written texts, where standard spelling is fairly general, and therefore individual words are recognizable, but regional and individual differences in speed and pronunciation are much harder to deal with. For instance, automated voice-recognition programs such as those used for finding long-distance telephone numbers deal with a limited range of pronunciations; many people with foreign accents need to go through the human operator. Also, different politeness conventions, conversational stragegies, and colloquial expressions which are actually metaphorical, all could get in the way of simply calling up people in different countries and expecting to converse as easily as with one's neighbours.
Bryn LaFollette said,
May 29, 2008 @ 1:51 pm
Honestly, no matter how fast computers get, this still won't give them knowledge about the world; something which is often crucial to correctly translating something that could be ambiguous in either the source or target languages. Some of the better statistical approaches come close to approximating this just due to approximate context, but this still leaves out common ground from discourse and other information which in some languages isn't always explicitly uttered, as in pro-drop languages or going from a language with no explicit number of gender specification to a language that requires it. These are the basic pitfalls that anyone in introductory Linguistics courses are going to learn about, but sadly this still isn't something that people widely understand, thinking instead that it's just that our computers aren't powerful enough yet.
When it comes right down to it, translation isn't easy, even for human beings! Just ask anyone that for the first time tries to reproduce a document in one language they know well into another language they know well. If it's hard work for a human, and never cut and dry. And if there's one thing that computers as they are today aren't good at, it's making non-deterministic judgement calls.
Ricky said,
May 29, 2008 @ 2:48 pm
I believe I remember reading that early computer scientists in 1960
in 1960 predicted good machine translation in 10 years, by
1970. 38 years overdue so far.
Lugubert said,
May 29, 2008 @ 5:04 pm
"Honestly, no matter how fast computers get, this still won't give them knowledge about the world"
That's a variation of what I claim: The more translation software we get, the more the demand for professional translators will be.
Computer programs might give you a general idea of the contents of a text, and the need if any for a _real_ translation. But for example for a manual that helps machinery users in not destroying themselves or their environment, or a leaflet that helps a patient use medication in a safe way, there's nothing like human pros who have the language skills as well as a profound understanding of the subject area.
Peter said,
May 29, 2008 @ 6:01 pm
Regarding the increasing trend towards regional and local linguistic nationalism: I believe globalization and the rise of global languages such as English or Chinese is a positive causal factor here. If you have a choice of either Catalan or Spanish for your children's mother tongue, but all your business communications are done in English, then you may as well choose Catalan for the mother tongue. In other words, the rise of English as a global language means that the pair (English, Catalan) is no longer less valuable than the pair (English, Spanish), while either pair is more valuable than the pair (Spanish, Catalan). Indeed, in multi-language countries such as Spain, the trend is self-reinforcing, since the pair (English, Catalan) is more valuable for Catalans engaged in inter-regional communications when people in other regions are choosing (English, Basque) or (English, Galician) over (English, Spanish).
john riemann soong said,
May 29, 2008 @ 7:56 pm
"thinking instead that it's just that our computers aren't powerful enough yet"
I suppose there's a difference between the power of an algorithm and the power of the hardware. I can foresee some pretty advanced heuristic techniques.
Several bots run on Wikipedia that have to do actual content-reading, to aid in catching vandalism, copyright violations, etc. Now the algorithms are pretty primitive, but context-judgment is not absolutely impossible for a computer — eventually however we must move past statistical engines to something with more powerful theoretical foundations.
No matter how much advanced math is used in statistical translators, they will never be the sort of "ideal translator." As it stands though statistical translators are more feasible than theory-based translators. I believe a similar issue concerns artificial voicing — currently what they do (IIRC) is take voice samples and attempt to construct phonetic realisations of words with them, rather than create phones from theoretical foundations. Currently when you attempt to do the latter, you get metallic robot-like realisations. But the latter must overtake the former sooner or later.
linda seebach said,
May 29, 2008 @ 11:10 pm
Mark said, "maybe I just should have sounded the tocsin for preservation of endangered languages."
I discover I'm not sure what "tocsin" means here. I have two overlapping senses having to do with bells. One is sort of like "knell," announcing a death. The other is "alarm," meaning we have to become active if this is to be accomplished.
Either could be true.
I have to confess, though, that in the original version I said, "meaning we have to prevent this," and I doubt that was Mark's intention.
marie-lucie said,
May 30, 2008 @ 12:24 am
I don't remember seeing the word "tocsin" in English, but in French "sonner le tocsin" refers to a special pattern of bell ringing used in earlier times in to warn the population to stop everything and take shelter (eg within the walls of the castle or city) as an attack was imminent. "Tolling the knell" as I understand it indicates that a death has already occurred.
Matthew Stuckwisch said,
May 30, 2008 @ 2:18 am
A small correction Mark, Catalonian, Basque, and Galician weren't made official by the European Charter, rather, by the Spanish Constitution of 1978.
Also, Peter, the trend here here in Spain has been a trilinigual approach. By law all citizens must learn Spanish, and in the regions were there is a coöficial language, all those schools will also teach that language under varies plans of study. Currently in the Basque country there are three options: Basque as primary language, Spanish as the secondary; bilingual program with everything half and half; and Spanish as primary with Basque as secondary; though this three-way system is set to change soon. In Catalonia Spanish is barely taught and their ability to understand and write it in the youth has diminished greatly. In any case, in many of these Spanish+local language schools, there are also other foreign languages taught, primarily French, German, or English, mostly the latter.
Mark Liberman said,
May 30, 2008 @ 3:03 am
@Linda Seebach: I only associate "tocsin" with the idea of an alarm bell, which is why I used it here. I didn't check, but now that I do, it seems that this is justified both historically and in contemporary usage.
The OED's first sense is "A signal, esp. an alarm-signal, sounded by ringing a bell or bells: used orig. and esp. in reference to France"; the second sense is "transf. A bell used to sound an alarm". No sense related to funerals is given.
When I check on Google News, all the recent examples of "tocsin" that I find are alarm bells.
However, when I check historical newspaper archives to look specifically for "tocsin" in connection with funerals, I do find a few uses of that kind. I think, therefore, you deserve a conditional unpology: I'm sorry if anyone was confused by my use of the word "tocsin".
@Matthew Stuckwisch: Thanks for the correction — I should have checked the history for Spain, as I did for France. (Though come to think of it, I knew that Catalan was already co-official when I first visited the area 25 years ago or so.) I'll stand by the main point, though, which is that the movement towards European union, starting in the 1950s, has increased rather than decreased the role of minority languages within the member states.
dr pepper said,
May 30, 2008 @ 3:17 am
Do Aragonian and Navarrese still exist?
John Cowan said,
May 30, 2008 @ 2:45 pm
Dr. Pepper: Ethnologue thinks not: the people in those parts of the country speak a local variant of standard Spanish. See http://www.ethnologue.com/show_family.asp?subid=90062 for a family tree of extant Ibero-Romance.
marie-lucie said,
May 30, 2008 @ 3:38 pm
In French "le tocsin" is not a bell, but a particular pattern of bell-ringing. In the Middle Ages and long afterwards, church bells (or in some places, city bells) used to be the main means of delivering fast news to an illiterate population. Churches which could afford it had several bells of different sizes and timbres, which could be rung in various patterns (as in the song about London bells and what they "say"). Different patterns of bell-ringing were associated with various types of announcements (the commonest being for announcing mass), and "le tocsin" was one of these patterns, used long and loud, raising terror in the population.
Stephen Jones said,
May 30, 2008 @ 6:37 pm
—–"In Catalonia Spanish is barely taught and their ability to understand and write it in the youth has diminished greatly."——
From which I gather you don't live in Catalonia.
The only non-Spanish speakers in Catalonia are European and African immigrants. The reason for the introduction of immersion in schools was demand from lower class Spanish speakers in the industrial areas of Catalonia. As Catalan was necessary for all regional and local government jobs in the region, as well as a fair number of jobs in the private sector, they were worried that their children would be disadvantaged if they didn't have an adequate knowledge of it. Their attitude was "you teach Catalan at school because we can ensure he speaks Spanish at home".
The opposition came from the entrenched Spanish speaking civil servants who had arrived before autonomy and the official recognition of Catalan.