Welcome / Bienvenue / Benvinguts / Bienvenidos
For information about my translation services, please visit the main site.
Pour des informations sur mes services, merci de regarder le site principal.
Para información sobre mis servicios de traducción, visite el web principal.

Conversion of Déjà Vu memories into MemoQ memories

If you export a Déjà Vu (DVX) memory or terminology database and import it into MemoQ, you lose some of the data such as the client, subject, project, user name and creation date. This is because the tmx format created by DVX does not match the tmx format created and understood by MemoQ. For example, Déjà Vu has separate creation dates and user IDs for the source and target, whereas MemoQ has a single creation date for a translation pair (which makes more sense). Also, the tmx created by DVX contains the subject and client codes, not the actual names. For example, if you used the subject “33 – Economics” in DVX, you will be importing the number “33” as the subject, not the word “Economics”. Similarly, if you used client codes, like “MST” for “Microsoft”, you’ll be importing the code rather than the full name.

Anglo Premier recently migrated from Déjà Vu to MemoQ. After much labour we succesfully converted our translation memories and terminology databases, preserving all the subject and client data and the dates. We initially described the process on this blog, but the procedure is complicated to follow and the script we created won’t run properly on all versions of Windows. It also requires the user to have Excel and Access 2003. Instead, we are offering to convert your translation memories and terminology databases for you. For a fee of €20 or £16.50 we will convert a translation memory or terminology database, and for €40 or £33 we will convert up to four databases. None of the content of your databases will be read and we will delete the databases from our system as soon as the conversion has been done and the file(s) have been sent to you.

If you wish to use this service, please contact us via the contact form on our main website.


Machine translation and context

A sentence I’ve just translated is an excellent example of the advantages and disadvantages of machine translation. The original sentence said this:

Los rankings se basan en indicadores sociales y económicos.

Google Translate offered this:

The rankings are based on social and economic indicators.

This is a good example of how machine translation can speed up the translation process. The translation is almost perfect. I say almost, because the translation doesn’t quite work in my context. The reader is left asking “Which rankings?”.

The original sentence is actually talking about rankings in general, rather than any specific rankings. Unfortunately Spanish does not make this distinction in the use of articles, so the word “los” is needed whether talking about rankings in general or specific rankings referred to earlier in the text. Google Translate works sentence by sentence, so it has no way of knowing whether the word “the” should appear at the beginning of the English translation.

Another similar problem comes up when I translate biographical texts. Imagine a sentence in Spanish that says the following:

Nació en Tolosa en 1960, pero desde 1970 vive en Roma.

Is Tolosa referring to the city of Toulouse in the Languedoc region (Tolosa is the traditional Spanish spelling of the city) or the small town in the Basque Country? OK, so I’ve deliberately come up with an ambiguous place name, but the other problem in this example does occur more often: is the text talking about a male or female? Google Translate has no way of knowing, since it only looks at the context of the sentence. It normally chooses a sex seemingly randomly. In this particular example it has produced a translation that does not specified the sex of the person:

Born in Toulouse in 1960, but since 1970 living in Rome.

Although it has avoided assigning a sex to the person the text is talking about, the translation is unacceptable and would need considerable editing. By changing the sentence slightly I can force Google Translate to assign a sex:

Nació en Tolosa en 1960, pero desde 1970 vive con sus padres en Roma.

Born in Toulouse in 1960, but since 1970 living with his parents in Rome.

The pronoun “his” is used, but the person we’re talking about could just as well be female. As additional evidence that Google Translate doesn’t use context I will now ask it to translate the following two sentences together:

Julia es una ilustradora francesa. Nació en Tolosa en 1960, pero desde 1970 vive con sus padres en Roma.

Google Translate provides the following translation:

Julia is a French illustrator. Born in Toulouse in 1960, but since 1970 living with his parents in Rome.

Google Translate still uses the word his, yet to any human translator it is blatantly obvious, thanks to the context of the first sentence, that the correct pronoun is her.