RFP Planning: 9 Ways to Tell if a Vendor Has a Mature MT Knowledge
Machine translation (MT) has entered a golden age thanks to artificial intelligence and other cutting-edge technologies. The MT market is undergoing constant transformation due to ongoing innovation – but the maturity of the technologies on offer varies drastically. How can companies with software translation needs reap the benefits of this technology to achieve the best possible price-performance ratio? Milengo shines a light on nine essential issues you should bear in mind when searching for an LSP that offers MT services.
1. Suitable content types
It’s important not to think of machine translation as a cure-all solution when it comes to translating your company’s documents. MT can work wonders with software documentation but might fall short when faced with stylistically challenging marketing material, for example. An experienced vendor won’t promise you the moon – instead, they will provide you with comprehensive advice on where it makes sense to use MT, and which content you’re better off leaving to a human translator.
2. Language combinations
An MT platform’s performance can vary greatly, depending on which languages it’s used to translate. It might achieve incredible results for software translations from English, while producing clunky German target texts – and it might not even support other languages. That’s why you should find out early on if your potential vendors offer premium quality MT for the language combinations you need.
3. Price transparency
Significantly lower costs are often the main argument put forward in favor of MT. When carrying out an RFP, you shouldn’t just settle for the vendor’s average prices for your required language combinations – you should request separate quotations for human translation and for machine translation post-editing (MTPE). That way, you’ll be able to compare directly the price difference between these services.
4. Defined quality levels
Correct usage of terminology, tone, and style are incredibly important when it comes to your company’s corporate language and software localization. You need to be able to count on MT meeting your high quality standards. That’s why you should ask the following specific questions: How exactly does MTPE differ from a conventional workflow made up of human translation and revision? Does the vendor offer different, clearly documented quality levels (such as “light post-editing” or “full post-editing”)? By clearly laying down the quality standards you need, you ensure that you won’t have to contend with any unpleasant surprises later on.
5. Post-editing expertise
Post-editing – or in other words, reworking a raw machine translation output to meet specific quality standards – calls for translators to have a different set of qualifications from those needed for conventional translation. The role of the post-editor is a relatively new one, and is still in the process of professionalization. That’s why in an RFP you should ask your potential vendors whether they have clear post-editing guidelines for software translation, if they apply comprehensive quality assurance methods, and if they thoroughly train their translators in the discipline of post-editing.
Modern MT systems are capable of “learning” – which means that they can be given company terminology and software documentation so that they gradually improve their results. That’s why you should select an LSP that can offer you a wealth of experience in training MT engines, or even has a range of MT solutions that can be optimized for specific content types or language combinations.
7. Data security
Free MT services, like the basic version of Google Translate, save your translated content on their servers and evaluate it to optimize their own algorithm. This is highly problematic in terms of data privacy, especially when it comes to sensitive business information. You should always ask an LSP that uses third-party MT engines if they have data protection measures in place that match up with your own compliance requirements.
The quality of machine translated text is continuously getting better – which means more fluent texts, better cost efficiency, and a shorter time-to-market for your translated software. But you’ll only be able to truly benefit from this technical progress if your LSP keeps up to date with the latest MT trends and incorporates them in its workflows. That’s why you should ask in an RFP: What level of maturity does their MT platform have? What MT innovations are they working on currently? Do they have an experienced MT department that keeps an eye out for new developments in this field?
9. Technical integration
Most companies use translation management systems (TMS) like memoQ, Trados Studio, or Across for processing translated content. But not every MT system is compatible with every TMS. Therefore you should always inquire whether and how machine translation can be technically integrated in your TMS or editorial system at the earliest stage possible of an RFP.