Deepl Online Machine Translation smarter than Google Translation

in #english6 years ago

As a translator I need a few tools to help me perform my work.

I started translating in 1982. At the time I mainly used specialised dictionaries and my own documentation, I collected here and there (various catalogues of spare parts and equipment for instance).
I sometimes had to go to Pompidou Centre Library in Paris when I had something more demanding.

Pompidou centre library patricklemarie steemit.png

When first machine translation software was available, I carried out a few unsuccessful tests that convinced me that translators‘skills were still much better. I was at the time really confident about my future, all the more as The Internet was not really allowing other translators located outside France to compete with me on account of the time it took to send a translation to the other side of the globe.
As technology improved in this domain, I regularly checked what I was offered, without really being convinced.

I switched to Computer Assisted Translation software in 2000, first working with Trados and then Wordfast which are translation memory programs, not automatically translating the source text but reusing what you have already confirmed as good translations, as you keep translating throughout your document or when reusing former documents you translated by indexing every source paragraph with every corresponding translated target paragraph (translation memories).Wordfast Computer Assisted Translation patricklemarie steemit.png

A few years ago I also started using Linguee which is a web service that provides an online dictionary for a number of language pairs. Linguee incorporates a search engine that gives access to large amounts of bilingual sentence pairs found online, especially from bilingual websites. I even found texts from my websites indexed by Linguee. It doesn’t mean Linguee always provides you with 100% accurate language pairs and the professional translator must still check the source of the suggested bilingual paragraphs. Most of the time I trust documents published by the European Union or the Canadian Government or other well-known institutions.
Linguee bilingual equivalents patricklemarie steemit.png

I also use IATE the European language database to cross-check sources.

IATE skipper patron pêcheur patricklemarie steemit.png

Until a few weeks ago I thought my translator life would go on as usual with probably a few technological improvements similar to Linguee but with no striking revolution.
However a month ago, a relative told me “Patrick I am going to show you something that is going to kill your business !”. DEEPL

I doubted his assertion, but as I am not the kind of man who is certain about everything, I decided to try my present (and future) competitor.
I put Deepl machine translator through a simple test as follows:

DEEPL test number one patricklemarie steemit.png

I carefully checked the translation from French into English, eager to find any error.
I did not find any. Honestly (as you will notice) the target paragraph was fully satisfactory.
OK my dear “translator robot” do you think I am the kind of buddy to give it up after the first attempt?
I found something a bit more difficult, a bit trickier. A paragraph written in French by one of my clients, a marine surveyor.
And the experiment went on as follows:

Deepl test number two patricklemarie steemit.png

I carefully examined my robot translator’s work, determined to find any weakness :
Source paragraph: “Suivant la déclaration de l’armateur, lors de la dernière marée au chalut, la capitaine a ressenti de fortes vibrations au niveau de l’hélice.
Le patron débraye sa machine et il constate qu’il vient d’engager son hélice par un filin non identifié à la dérive entre deux eaux.”

Deepl Translated target paragraph:
"According to the shipowner's statement, during the last trawl trip, the master felt strong vibrations at the propeller.
The skipper disengaged his engine and noticed that he had just engaged his propeller with an unidentified line drifting between two waters."

I examined how my translator buddy had translated “dernière marée” which is not that easy to translate if you don’t take the context into account.

“Dernière marée” in this case doesn’t mean “last tide”.

My machine translator buddy chose “trawl trip” which is correct when taking the context into account.
Ok, Ok wait a minute …

For “Capitaine” it translated “Master” which is also correct.

However the French marine surveyor used a synonym in the second paragraph to refer to the “Master” (le capitaine), “Le patron” also meaning the “Master” in English, and this time my robot translator (my translator buddy) chose the term “Skipper”, which is also correct when you refer to the master of a fishing ship, even though “Skipper” is often used to refer to the sailors competing in sailing races such as Transat Jacques Vabre departing from Le Havre, for instance.

You got me my dear robot translator; you son of …. Artificial Intelligence!

When I think you were born in July 2017. At your age I could not even speak my proper mother tongue correctly!!

Have you ever wondered what your future is going to be thanks to or because of Artificial Intelligence?


Please don't use fr tag without fr text ;-)

j'ai mis le tag fr car je parle aussi d'un outil d'aide à la traduction dans le couple de langue français/anglais, qui peut aussi aider les francophones.

C'est vrai. Je viens de lire ton article. Cela faisait plusieurs mois que je n'étais pas venu sur Steemit. C'est bien aussi d'avoir plusieurs avis n'est-ce pas ?

Cela ne retire en rien l'intérêt ni le mérite de ton article. Un ami m'a parlé de Deepl il y a environ Trois semaines. La plupart du temps j'ai les plus grands doutes concernant ces outils de traduction automatique.

Deepl reste quand même un outil à utiliser avec discernement, et chaque texte traduit doit quand même être relu par quelqu'un qui possède au moins les compétences de Deepl !

J'ai fait d'autres tests depuis cet article, notamment pour une traduction d'un contrat. Un traducteur qualifié demeure encore indispensable pour ne pas éviter des erreurs possibles d'interprétation.

Oui bien-sûr ! A bientôt ;-)

Coin Marketplace

STEEM 0.34
TRX 0.12
JST 0.047
BTC 70496.64
ETH 3495.63
USDT 1.00
SBD 4.65