Machine translation in the European Commission Josep Bonet - - PowerPoint PPT Presentation

machine translation in the european commission
SMART_READER_LITE
LIVE PREVIEW

Machine translation in the European Commission Josep Bonet - - PowerPoint PPT Presentation

Directorate-General for Translation Directorate-General for Translation Machine translation in the European Commission Josep Bonet LangTech 2008 Rome, 28/02/2008 DG Translation European Commission Stating the obvious Directorate-General


slide-1
SLIDE 1 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Machine translation in the European Commission

Josep Bonet LangTech 2008 Rome, 28/02/2008

slide-2
SLIDE 2 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Stating the obvious

DGT means DG for Translation Machine translation is about Translation Therefore DGT is concerned with machine translation Eh… Maybe, but not sure

slide-3
SLIDE 3 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

The plan

The infancy (how it all started) The first maturity (translators get focus) The middle age crisis (I don’t get no satisfaction) The second maturity (will this time be for real?)

slide-4
SLIDE 4 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

How it all started

When? In the mid-70s By whom? DG XIII, Telecommunications, Information Market, and Exploitation of Research Why? To overcome language barriers Why? To support the development of enabling technologies What? Rule-based system for EN=>FR and FR=>EN

slide-5
SLIDE 5 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

How it all started (cont’d)

Initial systems improved Dictionaries enlarged and completed Rules added and improved New systems: German, Spanish, Italian, etc. Joint venture DG XIII – SdT

slide-6
SLIDE 6 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

The second w ay

Eurotra

Big research project Many teams in most of Member States Innovative approach Too heavily dependent on computing power No working MT system delivered Big boost to language industries in Europe

slide-7
SLIDE 7 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

The plan

The infancy (how it all started) The first maturity (translators get focus) The middle age crisis (I don’t get no satisfaction) The second maturity (will this time be for real?)

slide-8
SLIDE 8 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Translators get focus

(< 1990) Translator equals monk or if I had a dictaphone Terminology and documentation go digital Typing kind of goes digital Translators scribble or dictate Translation is an art and the translator the artist

slide-9
SLIDE 9 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Translators get focus (cont’d)

(> 1990) PCs become pervasive Translation is (almost) fully electronic Tools for translators get the focus Art versus consistency. A new view on quality Time is gold. And gold matters

slide-10
SLIDE 10 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Translators get focus (cont’d)

New orientation in the ‘90s EC to finance research projects, not to develop Help the industry become strong Stress on pre-competitive research End-user products out of scope

slide-11
SLIDE 11 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Translators get focus (cont’d)

(Mid-’90s) Development of MT transferred to DGT Development of further language pairs Improvement of dictionaries and lexical routines Total number of pairs available: 26 English and French as source or target language Coupled (to and fro) with DE, IT, ES, NL, PT, EL, SV, DA

slide-12
SLIDE 12 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Translators get focus (cont’d)

Varying quality From rather acceptable To rather unacceptable (prototype) For best output, less important text, easier input and more committed community: MT with light and quick post-processing Keep you eyes on the client Tam-tam and Tchernobyl

slide-13
SLIDE 13 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

How do you get MT in the EC?

1) E-mail > phased out 2) Euramis: integrator of language services 3) Web interface 4) Automated pre-processing Methods 2 and 4 only for translators Methods 1 and 3 also for all EC staff

slide-14
SLIDE 14 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

The plan

The infancy (how it all started) The first maturity (translators get focus) The middle age crisis (I don’t get no satisfaction) The second maturity (will this time be for real?)

slide-15
SLIDE 15 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

I don’t get no satisfaction

Creating ex nihilo or re-using? The green era: recycle rules Europe enlarges, so what? 23 languages > great impact Redefinition of translation business Expandability of systems The corpus is king TM won the battle against MT

slide-16
SLIDE 16 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

I do get satisfaction, by the w ay

There is life beyond translation Who are the big users of MT in EC? Are they satisfied?

slide-17
SLIDE 17 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

The plan

The infancy (how it all started) The first maturity (translators get focus) The middle age crisis (I don’t get no satisfaction) The second maturity (will this time be for real?)

slide-18
SLIDE 18 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Will this time be for real?

Fully Automatic High Quality Translation is out Fully Automated Usable Translation is in Corpus-based developments and global markets Quality is in the eye of the reader The user finally gets the focus

slide-19
SLIDE 19 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Caring about the user

Translator: prefers quality, even if incomplete End-user: requests all the information, even if little readable Don’t overlook tolerance thresholds Language is extremely personal It involves deep feelings: sensitive issue MT should be seen as an added service, not a denial of service

slide-20
SLIDE 20 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Let’s start w rapping up

Availability of big corpora has boosted developments DGT has released its corpus of aligned EU legislation in force in 22 languages Exponential growth of demand for quick and cheap translation User satisfaction with MT solutions Resignation? Rather imperfect than non existing The case of Microsoft’s Knowledge Base

slide-21
SLIDE 21 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Let’s start w rapping up

Concentration on major languages Most European languages have few speakers MT remains mostly intelligence-related

slide-22
SLIDE 22 European Commission

DG Translation

Directorate-General for Translation Directorate-General for Translation

Thank you