Terminology checking

by Steven Marzuola

As careful as we try to be, there are times when we make mistakes. Therefore, I take every opportunity to find a way to check my work.

One very useful tool is the terminology check feature built into my translation memory software. From the DV 3.0 help system:

Check terminology searches the project lexicon and the terminology database for words and groups of words in source text and makes sure that the correct translations have been used in target text.

The problem is: as a TDB gets bigger (and more useful), it presents a problem as a quality reference.

1. In some cases, the TDB will have multiple translations, for different subject and clients. Most of the time the Spanish word “auto” means “automobile”. But in a legal document, it could be translated as a “brief”. Both translations are included in my TDB. The client and subject attributes usually help the program select the translation that is suited for the current project. But the terminology check will find the term I didn't use, and mark it as an error.
2. When translating into Spanish, adjectives agree with nouns. So the English adjective “ferrous” could be translated into Spanish as “ferroso”, “ferrosa”, “ferrosos”, or “ferrosas” depending on gender and number. To handle this, my TDB has the pair “ferrous - ferroso_”. The underscore character makes me notice that agreement is needed. Good for quality. But bad for the quality checker, because any of the correct words will be flagged as incorrect.
3. Say that my source document includes a company name: “ABC Technologies”. Normally a company name is not translated. But the TDB has the pair, “technologies = tecnologías”, and if I leave the name in English this will be noted as an error.

Those are just a few examples of conflicts. A word that is correct for the current context will be wrong because the TDB has another option . As it grows, the TDB becomes more and more useless as a tool for checking because too many errors will be found.

So I don't use my regular TDB. But I still need to run a terminology check.

One solution is to add the words that are important in the current project, to the Lexicon.

Another type of text to be checked is strings of letters that are not words. These are things like codes, part numbers, model numbers, etc. For example, my source documents may include alphanumeric strings, such as:

TR-02-SA
DR-H-221

My English-to-Spanish “check” TDB includes the following:

• all single letters that are not words. “A” and “I” are words in English with Spanish translation, but almost any time my source document has the word “B”, the Spanish version should also have it. Same for  “C”, “D”, etc.,
• all two-letter pairs that are not words or abbreviations: AA, AB, AC,... BA, BB, ... up to ZZ. If my source document has the letters “AB”, so should the translation.
• the numbers “two” to “forty”.
• Roman numerals “II” to “L” (50)
• the months of the year: “January” to “December”. Also, their three letter abbreviations: “Jan”, “Feb”, etc.
• other terms, including abbreviations and acronyms, that should be either (a) left untranslated or (b) have a preferred translation practically 100% of the time.

Together with the numbers check, this “check” TDB has saved me from errors on many occasions. I recommend it. If you would like some help setting up a check TDB for your language pair, please contact me.