I was asked this question today.
It wasn’t the first time. If I’m honest, it annoyed me that I should have to answer it at all. But I guess if you don’t work in the language industry, you might perceive Google as a trustworthy company who can do no wrong, so you could be forgiven for thinking that their machine translation would be equally reliable. I’m answering it here on the language blog, to share with anyone who may be guilty of having the same thoughts.
It’s surprising (to me, at least) how many times I hear things like;
- So basically you do the same as Google Translate?
- Why should I pay you anything when I can get Google Translate to do it for free?
- Do you use Google Translate for all your translation?
- Do you just have one big computer who does all the translation?
(the answer is NO to all of the above)
Over the years I’ve lost quite a few projects to people translating legal contracts, user guides, websites and most recently, translating food packaging, with Google Translate. How do we compete with something that’s perceived as being free? I can talk about how much time, effort and money is required to get the original English document right, so surely when approaching the translation aspect, at least some respect should be paid to the quality of language used, especially when the originator may not be able to read the results. But not everyone listens.
The practice of maintaining a multinational marketing message and ensuring the feel of the original language is kept the same takes a high degree of knowledge and skill that can only come (at least for the moment) from human involvement. Machines are unable to accurately convey or understand all the nuances involved with a localization project. When companies rely on this type of machine translation, the shortcuts become painfully obvious to readers of the final product. One of the most popular posts on the language blog is about when packaging translation goes wrong.
Yeah, but it looks Right
What really makes me sad is when we see examples of translation companies using Google Translate as their own work. Whenever I spot this (believe me we can tell) I do my best to tell the company involved.
Earlier this year we used this very method to remove a translation agency supplying a very large global retailer with translation to be used on their international packing. If you’re trying to remove an incumbent company a great way of getting the client to listen to you is for you to proofread their text and present the findings to them, especially if it has a legal connotation to it (for instance food packaging or medical documentation).
The translation house in question claimed to use professional translators (aligned with EN15038 which means a translator, reviewer and an independent QA is employed to check the target document translation) and yet we found examples of text which had been translated using the same output Google Translate spits out.
Let me give you an example.
Google translate (image taken 27th November 2013) returns the French translation of ‘may contain nuts’ as ‘Peut contenir des noix’ – looks about right doesn’t it? Well noix translates as walnuts so the phrase returned can be understood as ‘may contain walnuts’. Are nuts and walnuts different? They are if your child has an allergy. And they are if you want to legally sell your product in a French speaking part of the world.
I took my translation team around a store in-country to see for themselves exactly what had happened. We found the error on 40% of packs. We now provide all of their language support.
Polluted Translation Memories
The problem is made worse by adding incorrect terms, such as these, to a translation memory without the help of a terminologist. They then get used to return ICEs or 101% matches (depending on your Translation Engine of choice). One problem phrase can soon find itself making its way on to almost all of your translated information, partly driven by the bizarre trend of not charging for, or checking 101/ICEs matches provided by large agencies. I’m all for saving money but with things this important to the end consumer extra care really needs to be taken.
Terminology Management can be used as either a cash generator or a barrier to entry for your competitors, both ways add value to your business. Making a success of it is largely dependent on demonstrating where you are adding value to your customer’s supply chain.
A study by Schuetz, Joerg and Rita Nuebel, “Multi-purpose vs. Task-specific Application: Diagnostic Evaluation of Multilingual Language Technologies” shows us that the cost of correcting a translation error increases dramatically as the process moves on (from Authoring to Maintenance). The value can be generated by managing the terminology for the client, not adding value to errors and demonstrating the value/money saved in the process.
There are numerous terminology tools available on the market to help your team manage this process but the tool has to be fit for the purpose intended. I’ve seen everything from dynamic systems running on top of the latest technology to Google spreadsheets with columns for notes/comments. The point is it needs to work and it needs to be built on a solid foundation, investment in expensive IT systems sometimes makes your problems worse.
Alignment with Stakeholders
Other stakeholders in the process will have vital input into the quality of the end product. Depending on what you’re translating, these people/departments could be;
- the client
- the supply chain
- the marketing team
- the artworkers
- legislation agents
- the end consumer
All will have an equal view about the translation, but some views are more equal than others. There are tools to help you to manage this such as stakeholder impact Vs influence 2x2s that help you to highlight the areas where your systems will need to be aligned first. This image (I didn’t invent the model but I use it and present about it) gives you a very quick view into how to manage the stakeholders as they fall into each box, it is obviously pointless spending a lot of time on someone who has very little power or influence over the project.
Where does Machine Translation fit into the Process?
We use machine translation. A little at the moment but we’ve invested a lot of $$$ into it this year and will be using it a lot more in 2014. But (as stakeholders are) machine translation systems aren’t equal. We use a bespoke system built on Statistical Machine Translation which in turn is built on specific glossaries, translation memories and past work we have completed for each client in each language pair. This is then translated, post edited and outputted to the artwork stage (or whatever is next in the process). Further reliance on leveraged technology puts an even greater importance in Quality Management, i.e. if the investment in terminology has been ignored (or taken verbatim without a terminologist’s check) then the problems our competitor had putting translation errors on food packing will only get worse.
Yeah but, Can I use it?
Google Translate has its uses, but read the data usage agreement and don’t rely on the output if you’re going to count on it to keep you/your company out of court. It is a subsidiary of Google. This means that it draws on the same technology as the regular search engine does. For languages, it cuts the sentence into segments which are then researched automatically on the web. The problem is that the web is an interactive tool, with the entire world as its supplier. Google Translate can’t draw solely from quality sources as these are often formalised language and doesn’t reflect consumer demand for the translation of popularised phrases. Furthermore, it knows no grammar, it fakes grammar through its segmentation rules.
The process of taking your message, breaking it down to its constitute parts, understanding the external influences in region, applying the relevant legislation and putting it all back together in another language remains a complex one. Don’t rely on a free service to do this (you wouldn’t for almost anything else in your business) and if your LSP is using it – talk to us.