skip to Main Content

Choosing the right Machine Translation option

For newcomers to the field, the choice of Machine Translation (MT) engines available is mind-boggling: rule-based, statistical, example-based, hybrid, multi-engine, system combination-based. We developers shouldn’t wonder why buyers are confused about the range of products and services available today.

Most of today’s MT engines – like those in Lingo24 – are statistical. Even ‘old school’ rule-based MT systems are likely to machine translationcontain statistical components nowadays. Freely available online MT engines such as Google Translate and Bing Translator are also statistical. For many users, MT is Google Translate.

For some use-cases, Google and Bing are fine. However, there are a number of reasons why users are likely to be better off having a customised engine built by a company like ours. In having our own industry-leading MT R&D team, Lingo24 directly controls the development of the engines we market; don’t confuse the few companies with real development capability with those who merely sell on this technology, often as a rebranded product.

Customised engine builds using a client’s own translation assets – translation memories (TMs) and glossaries – considerably outperform freely available web-based MT offerings in almost all cases. There are security risks when using freely available online systems, and data leakage from company employees’ use of these systems is a major concern; read the small print, as whatever you upload on these sites for translation can be used by the providers of that service for their own benefit.

Furthermore, freely available online systems are general-purpose tools that have to be able to translate any input whatsoever; they cannot be customised to the precise requirements of a client. In contrast, Lingo24’s bespoke MT engines are combined with our in-house TM technology to obtain the best of both worlds for our clients: TM leverage for high-scoring fuzzy matches, and first-class MT technology. We guarantee data security for our clients: customer-supplied translation assets will only ever be used to build engines for that client. We are happy to delete translation data on completion of a job, if required.

In addition, as more and more of the client’s text is pushed through a customised engine, and post-edited by our team of experienced linguists, more MT training material becomes available and the quality of the engine improves. There’s no guarantee that you’ll ever see any improvement in quality with freely available systems.

Lingo24’s MT developers have a wealth of previous experience in developing top-quality engines for a wide variety of language pairs and domains. We can build state-of-the-art engines for any language pair. For clients requiring 100% adherence to the contents of user-supplied glossaries, we can guarantee that source-language matches occurring in the glossary are translated ‘as is’. Our solutions can be integrated with TM, or users can interact directly with their engines via our API.

In sum, we are confident of the quality of the MT engines we build, and we especially encourage clients new to MT to engage with us. The first interaction with an MT provider is all important, and the advice you get from Lingo24 will not be bettered by anyone. If you then choose to engage with us – and our entry-level offerings are attractively priced – the level of engagement will be second-to-none. Come and join Lingo24 and you won’t look back!

Andy Way

Professor Andy Way is Lingo24’s Director of Machine Translation. He has more than 25 years’ experience in the field, building a world-leading research group at Dublin City University. He is currently President of the International Association for Machine Translation and the European Association for Machine Translation, as well as Editor of the Machine Translation Journal, the leading journal in the field.

Back To Top
×Close search