Before you read the story, please accept my apologies for a missing screenshots. My customer does not want me to show his job in any way.

I did a small job (3000 words) and as I knew the customer, I just made a quote using a word count tool on my phone. Turned out the job was way more difficult than I thought it would be. I spent more than 2 days on this small job…

As I’m co-owner of a company that created a tool to make a source analysis, I feel very bad about this. So I checked if our tool (Fair Trade Translation) would have seen this job was going to be trouble.

Our translation tool only supports English into FIGSP for now, but I could have used the source text analysis as my source was English anyway. This is what Fair Trade Translation saw when it was too late:

It should have been “Automotive”. Our system not being able to tell what the text is about, should have been a first warning.

I know this estimate is not always right, because it is based on various source text properties and on machine translation performance (we compare how machine translation engines perform on the text), and that doesn’t tell something about my own skills as a translator. But when our tool says it will probably be difficult … I should have taken a closer look.

Many long sentences. Maybe the writer was not a professional writer. There were indeed a lot of long sentences with complex subclauses.

I also checked the Suggested Terms (the list I cannot show you – once more I’m sorry). That is a list of terms extracted from the source but the terms are also translated into context. As Dutch translation is not yet supported, I hardly ever use this feature myself. But it is good to take a look at the list anyway: about one third of the extracted terms was not really what I would call a term – so be it – but another third were terms I did not know at all… Half of those were not in any free online dictionary — Good to see that Linguee, a service we link the extracted terms to, did find at least some partial hits for some terms.

The Suggested Terms list could have been the main reason for me not to accept the job, or to the raise my quote. A pitty I didn’t check it before submitting it.

Last but not least: the text-reuse… Almost no reuse, even not on subsegment level. Every sentence was indeed hard work…

The Fair Trade Translation analysis was right, and it probably won’t be right all the time. After all, it is also based on NLP technology — we all know translators still out-rank machines by far. But if the tool indicates, in one or more reports, that a job isn’t low hanging fruit, a more careful approach may be very rewarding.

Another lesson learned…

(Also read: Back to basics – how does it feel to translate naked?)