Sadly, the primary objective for an LSP’s machine translation department is usually mostly reactionary: to support and retain clients whenever they decide to replace humans with MT. This is to say that, historically, the decision to use machine translation comes from the client. We’ve written about this before, and so have others, so we don’t need to go into too much detail here.
LSPs usually prefer to own all of the client’s language services needs, or at least as many as the client will trust them with. If some of these needs can be fulfilled with a machine rather than human translation, it is better to keep this part in-house, and leave no space for the competition. This means that most LSPs will gladly adopt machine translation as part of their process, but usually only as a reaction to pressure from losing their clients.
The secondary objective for MT departments is business development: to translate more with current clients and to find new clients for MT solutions. Salespeople (or solutions architects involved in the sales process) who are familiar with machine translation can better advise clients what machine translation is and how it can be used to improve the translation process. It’s quite the sales pitch.
In today’s technology-driven world, clients don’t want to hear about how to translate. Clients stopped buying translation years ago. They buy solutions that are driven by technology that seamlessly includes considerations for their international markets into their existing workflows without additional overhead.
If you are an LSP that is going to remain relevant, there are certain things to keep in mind for both production as well as sales.
The first step to diving into machine translation is making sure that you are set up to handle the job. LSPs don’t need to develop their own custom engine to start working with machine translation. In fact, we advise against it. Developing such engines is an ongoing investment that takes way more money than any LSP will ever see on a return. There is just no way for an LSP with its limited resources, to compete against the giants like Microsoft, Google, Baidu, and now Amazon and Apple.
What LSPs can do though, is make sure that they have some basic things in place:
When a client adopts machine translation, salespeople are faced with a choice:
Do they get intimidated and simply move on to the next lead, feeling that they couldn’t possibly deliver any innovative technology solution? Or do they jump in head first, selling not only machine translation management, but also up-selling engine training services and post editing, bringing in contracts worth millions?
The answer to this question is unfortunately out of the hands of the salesperson. Each salesperson needs to recognize the limitations of their own technology department. Unless they are confident (or at least optimistic) that their teams can deliver, they may tend to pass on potentially lucrative deals.
For those optimistic salespeople out there (and really, how many salespeople do you know that aren’t optimistic?), you will need to have some basic tools in your toolbox before selling in the age of machine translation:
For those that still have an unquenchable thirst for information about technology and machine translation, there are many good resources available. Of course, Nimdzi is always happy to share our insights so feel free to reach out if you would like to discuss further. But you don’t have to take our word for it.
There are many different opinions about machine translation floating around the interwebs. Some of these opinions are well informed. Some of them are nonsense. Some of them are presented well, and some of them read like an AP Calculus textbook. But if you have time, check out the information available from these two resources below.
*More information on “Upstream”, a book by Luigi Muzii with a forward from Kirti Vashee, can be found here.
Some machine translation providers are holding out hope for MT systems that adapt to document context. Could this development eliminate the need for custom MT engines? Will context-enabled MT help MT achieve human parity? Will we still need to customize a few years from now? Let’s discuss further.
5 August 2020