Taxpayers' money for subsidizing machine translation?
Thread poster: Maria S. Loose, LL.M.

Maria S. Loose, LL.M.  Identity Verified
Belgium
Local time: 19:16
German to English
+ ...
Mar 23, 2015

A couple of hours ago we were asked to sign a petition asking the European Commission to subsidize the development of machine translation. All of a sudden this thread has mysteriously disappeared. The poster probably asked site staff to remove it.
Nevertheless I still think it is an interesting question. I'm therefore starting a new thread.
The petition we were asked to sign is here: http://multilingualeurope.eu/#fullletter
In my answer to the original poster I stated that the European Commission should subsidize communication between European citizens by using taxpayers' money for training translators and for subsidizing communication via human translation instead of subsidizing machine translation. At a time of high unemployment in Europe it's scandalous to subsidize artificial intelligence. Any thoughts?

[Edited at 2015-03-23 18:07 GMT]

[Edited at 2015-03-23 21:53 GMT]

[Edited at 2015-03-23 21:58 GMT]

[Edited at 2015-03-23 21:59 GMT]


Direct link Reply with quote
 

dianaft  Identity Verified
United Kingdom
Local time: 18:16
Member (2013)
German to English
+ ...
Is there really a communication problem? Mar 23, 2015

Is there really a communication problem?

With the exception of the Brits, most Europeans under 45 speak more than one language at a conversational level. Language skills increase career prospects even in entirely unrelated industries, so students tend to take their foreign language lessons rather seriously. In Germany, you have to learn two languages (at least at a basic level) to gain your university entrance qualification. There are a range of interschool competitions at national and European level to promote the interest in foreign languages.

If translation was subsidised, how much of that subsidy would actually end up in translators' pockets? What would stop translation agencies from claiming those subsidies while continuing to outsource the work to India?
I hold my own views on unemployment which are largely not very pc and contribute nothing to the question of subsidies.

If there was subsidies for the enhancement of communication between citizens in Europe (slightly different to European citizens), the main area of focus should surely be that which affects us on a daily basis 'as citizens'. I would hold that this is the conversational competence of the language spoken at the place of residence. While there are courses for speakers of other languages and these are often offered for free, they can be very difficult to access in some areas and this certainly needs to be developed.

As to MT - it will be developed either way. If it is developed in India or Singapore, it will still affect us. There simply won't be many Europeans involved in that research and development. But I agree, money can be spend in a better way.



[Edited at 2015-03-23 19:07 GMT]


Direct link Reply with quote
 

MO-TS  Identity Verified
Netherlands
Local time: 19:16
English to Dutch
+ ...
Original thread Mar 23, 2015

http://www.proz.com/forum/business_issues/283559-petition_open_letter_to_the_ec_requesting_to_address_the_multilingual_challenge_in_their_forthcom.html

Is this the original thread you are referring to? If so, it did not disappear.


Direct link Reply with quote
 

Maria S. Loose, LL.M.  Identity Verified
Belgium
Local time: 19:16
German to English
+ ...
TOPIC STARTER
OFF TOPIC Mar 23, 2015

Thank you MO-TS. The original thread was moved to OFF TOPIC. That's why I couldn't see it.
Dianaft, they say in the open letter that more than 40 per cent of EU citizens never shop online in a language other than their native language, and I think it's true. Even outside the UK, a lot of people aren't proficient enough in a foreign language to shop online. Think about Eastern and Southern Europe. This summer I fell sick in southern Italy and needed urgent medical help. None of the physicians spoke a word of English and there were no translators or interpreters around. It was horrible.
The originators of the petition are claiming that private companies are developing machine translation only for languages profitable to them, i. e. big languages. They therefore want the Commission to subsidize language technology for languages of lesser diffusion, which are not profitable for private companies. In my opinion this money is better spent for University courses teaching languages of lesser diffusion and translation techniques and thereby creating jobs for human translators. Or it could be spent by reimbursing translation costs to, say, a small Romanian company which wants to have a website in Hungarian, so that Hungarian consumers can read it. It's probably best not to subsidize any of this, but we must not forget that tuition-free University studies are in reality a form of Government subsidy. What really annoys me a lot is that government money is spent for research into artificial intelligence that makes humans superfluous. If private companies want to take that risk let them do it but don't let the taxpayer pay twice: first for subsidizing this kind of research and then for paying unemployment benefits to those made redundant because of this technology.

[Edited at 2015-03-23 22:47 GMT]

[Edited at 2015-03-23 22:54 GMT]

[Edited at 2015-03-23 22:57 GMT]

[Edited at 2015-03-23 23:07 GMT]


Direct link Reply with quote
 

Maria S. Loose, LL.M.  Identity Verified
Belgium
Local time: 19:16
German to English
+ ...
TOPIC STARTER
Mar 23, 2015



[Edited at 2015-03-23 22:40 GMT]


Direct link Reply with quote
 

Jeff Allen  Identity Verified
France
Local time: 19:16
Member (2011)
Multiplelanguages
+ ...
where are MT systems developed? Mar 24, 2015

dianaft wrote:
As to MT - it will be developed either way. If it is developed in India or Singapore, it will still affect us. There simply won't be many Europeans involved in that research and development.


@Dianaft, this seems to be just assumption that MT development is or would be conducted in India or Asia-Pacific countries.

Several months ago I wrote up a report, based on facts on over 200 translation tool providers taken from published information found about them in the public domain over the past 15+ years, which instead proves the opposite case. This also refects market watch survey work which I conducted on behalf of the European Commission including an analysis of hiring trends in the area of natural language processing/computational linguistics.

1) India: To my knowledge, only 2 requests about MT have ever been made with regard to potential development resources in India, and both of those requests were initiated by the academic community, not business/industry.
2) As for Asia-Pacific Rim countries, I conducted a special project in 2000-2001 to survey all MT products for Asian languages in order to consider licensing opportunities. The result of that survey work and interviews showed that some MT products were developed in-country (with local management in the same country) for their home language in combination with only one or two other languages, yet the majority of all multi-language MT systems were developed in Europe and North America.
There are only a few exceptions.

The landscape of the translation software development community has not changed much since. And again today there are only a few exceptions.


Direct link Reply with quote
 

Jeff Allen  Identity Verified
France
Local time: 19:16
Member (2011)
Multiplelanguages
+ ...
commercial companies doing MT for less prevalent languages Mar 24, 2015

Maria S. Loose, LL.M. wrote:
The originators of the petition are claiming that private companies are developing machine translation only for languages profitable to them, i. e. big languages. They therefore want the Commission to subsidize language technology for languages of lesser diffusion, which are not profitable for private companies.


Maria,
if this is the main claim of the petition, then it is founded only on partial truth. Back in 1998 at the AMTA conference in Philadelphia, a panel session led by Laurie Gerber "The forgotten majority: neglected languages" made statements about the cost of creating a new MT language pair for lesser widely used languages. I was already working specifically on that topic and the so the panel session caught my attention.
Several years later (2006), the same financial figure was stated by an MT company to a potential customer as the amount needed to kick off and create an MT language for an Eastern European language. I was present at that meeting in a decisional role at that MT provider company.
However, this mainly reflects the cost of creating rule-based (grammar rules + general and topic-specific dictionaries) MT systems. I provide a 5 minute overview "for translators" of the several main types of MT systems at this link.
Language Technologies and Services
http://www.youtube.com/playlist?list=PLdABKB0CMCgbjdz5Fj62Xufshyzw8N1t2

However, the technology type started changing just about the same time. I was working on some of the first statistical MT systems back in 1998, and specifically on less widely used languages, funded primarily by government agencies to create rapid-development and rapidly-deployable speech-to-speech translation system for humanitarian relief and military check-point situations. About 4 years later, one company was the first to make its statistical MT (SMT) approach available to the commercial market. Now there are many of them doing the same. The SMT approach has made it possible to capitalize on language data assets (for example, TMs and term bases) to quickly train an SMT system on languages other than FIGES, Asian languages and Arabic.
Several commercial companies have indeed created different MT language pairs, based on an SMT approach, for such languages. An example of this is shown in the video of my invited talk at Microsoft headquarters in 2010 explains how I championed the Haitian Creole disaster relief translation technology project during the aftermath of the Haiti earthquake to put Haitian Creole into MS Bing Translator, Google Translate, and many other providers. The video is at the same link as the one provided above.

Over the past 5 years, there has been a significant increase, across a variety of MT providers, in the number of languages offered, when you compare it with the results of the survey work that I conducted back in 1998-2001 on this topic. Webinars and conferences talks are regularly presented by these companies on how new "exotic" language pairs are being implemented by different user institutions based on an SMT based approach.

The problem is not the technology, but rather with regard to the availability of data. I wrote quite a few articles and presented talks on this specific subject which call all be found in my slightly outdated list of publications and conferences talks. (2 Aug 2013).
jeffallen-all-publications-02Aug2013.pdf
https://app.box.com/s/1hyr0kj0rmr44jke0v12
Note: I have been planning to update this list with much completed and presented in 2014.

In just a few weeks, I will be presenting a talk about Minority language translation in general (see link below). This will not only address points about language technologies, but will cover wider range of relevant topics which are very relevant to the overall topic (including networking, trust/confidence development, partnership sustainability, non-profit NGO activities, pro-bono work, etc)

ELIA Networking Days Lyon 2015
17 April 2015
Making minority language translation a profitable translation business model
http://www.elia-association.org/session.php?p=27&s=26
Jeff Allen
Speaker biosketch:
http://www.elia-association.org/speaker.php?p=26&s=36


Direct link Reply with quote
 

Phil Hand  Identity Verified
China
Local time: 01:16
Chinese to English
Don't like MT =/= don't want MT Mar 24, 2015

I can't really agree with Maria on this. I don't like MT, as in, I don't think it's a useful technology yet for people who need serious translation done. But MT as an idea is awesome. Do we really want Europe falling (even further) behind in technology out of some Luddite impulse to protect translators' jobs?

I don't know enough about the politics of the particular petition to know if it's a good idea or not, but in general I think there should be massive research into MT and AI in Europe.


Direct link Reply with quote
 

DCM Linguistics  Identity Verified
United Kingdom
Local time: 18:16
Portuguese to English
+ ...
Can't agree, Phil Mar 24, 2015

We should always look into the politics behind every single petition we are invited to sign, especially when commercial entities are involved. The pretext that they use 99% of the time - that they're fighting for the underdog - is just that, a pretext. The end goal is always actually something else. (To give you an example, look into some of the TWB's practices).

And why would I want MT developed for Romanian, for example? So that colleagues go from being paid €0.04 a word to €0.01 a word? Is that your vision for our future as a profession, Phil?

I am also of the opinion that in fact there's no communication problem in Europe. Most of us have been graduating school speaking 2-3 languages, to one extent or another for generations now.

IMVHO, EU money can go into better projects.



[Edited at 2015-03-24 16:48 GMT]


Direct link Reply with quote
 

Gordana Podvezanec  Identity Verified
Croatia
Local time: 19:16
Member (2003)
German to Croatian
+ ...
MT and the rates Mar 25, 2015

"- Multilingualism is the key to European unity.

If people decide that they want to continue working without any TM - they may very well do so but I am advocating for the technological development in the area of translation.

The other issue comes with the rates that go lower due to TM repetitions etc....
I believe we should speak for the dignity of the profession, its professionalisation, decent rates throughout the EU and not fight against the technology.

[Edited at 2015-03-25 12:36 GMT]


Direct link Reply with quote
 
Peter Holozan
Slovenia
Local time: 19:16
MT equals SMT? Mar 26, 2015

Jeff Allen wrote:
However, the technology type started changing just about the same time. I was working on some of the first statistical MT systems back in 1998, and specifically on less widely used languages, funded primarily by government agencies to create rapid-development and rapidly-deployable speech-to-speech translation system for humanitarian relief and military check-point situations. About 4 years later, one company was the first to make its statistical MT (SMT) approach available to the commercial market. Now there are many of them doing the same. The SMT approach has made it possible to capitalize on language data assets (for example, TMs and term bases) to quickly train an SMT system on languages other than FIGES, Asian languages and Arabic.


I agree that SMT can quickly achieve machine translation for any language provided you have some parallel corpora to work on.

But on the other hand I don't believe that current state-of-the-art SMT is the quality of MT we want. It is good enough to get a gist from original but it cannot be mistaken for real translation (or to be relied on for shopping (terms of use, which can be diffucult to umderstand even in native language)). My language (Slovenian) for example, is morphologically quite rich and some information is also conveyed with word order (emphasise, which part of information is important). And in the last 5 years Google Translate for Slovenian doesn't show much progress.

So in my opinion, to get better MT we will have to move beyond SMT, which will most probably require a lot of linguistic data for the involved languages. And that can be a problem for less researched languages.


Direct link Reply with quote
 


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

Taxpayers' money for subsidizing machine translation?

Advanced search






memoQ translator pro
Kilgray's memoQ is the world's fastest developing integrated localization & translation environment rendering you more productive and efficient.

With our advanced file filters, unlimited language and advanced file support, memoQ translator pro has been designed for translators and reviewers who work on their own, with other translators or in team-based translation projects.

More info »
BaccS – Business Accounting Software
Modern desktop project management for freelance translators

BaccS makes it easy for translators to manage their projects, schedule tasks, create invoices, and view highly customizable reports. User-friendly, ProZ.com integration, community-driven development – a few reasons BaccS is trusted by translators!

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search