Today we have a guest post from Wordbee, a company based in Luxembourg that is developing translation software, including CAT tools and MT solutions. In this article, they give us some information about some the different MT options that are available on the market, including their own solution. We hope you find this information useful. Remember, however, that Machine Translation is not always useful. If you would like to use MT in your localization procedures, do not hesitate to contact us and we will advise you on which can be the strategy most suitable to you.
According to Google, Google Translate is used to machine translate approximately the same amount of text as 1 million books… in just one day (2012). But despite widespread application on the web and to some extent in the professional industry, advanced machine translation techniques like training are still left untouched by many Language Service Providers and some of the largest enterprises. Lots of businesses are sorely in need of leveling up.
The benefits of machine translation + post-editing to the professional industry are widely recognized, because costs are lowered and output is raised, both of which are necessities in a market that demands a huge number of words be translated every day.
Post-editing workflows, when integrated with a translation management system, are truly elite according to present-day technology.
What is training?
Training is when you post-edit your machine translation, and then send the results back to your machine translation system so it can “learn”. Training works when combined to statistical engines like Microsoft Translator or Google Translate, because those systems learn when more data is introduced. Rule-based systems, on the other hand, need better rules to learn.
Once a service provider or enterprise is training their own machine translation system, they improve their future machine translations with an essentially cost-free investment. That’s smart, and truly on the cutting edge.
Integrating machine translation to web-based systems via API
A web-based translation management system like Wordbee is a crucial step toward an ultimately pain-free workflow, because:
- Machine translation occurs within your own system via an API connection to your machine translation software.
- Translation memory can automatically be leveraged prior to machine translation.
- Because post-editing occurs within the web-based system’s CAT tool, analytics data can tell you things about the quality of your machine translation.
- The entire process from the customer order to the delivery can be automated, including the assignment of post-editing and DTP jobs.
- Closing the loop by training your own machine translator is easier than ever.
- Because translators use your browser-based CAT tool, you can work with translators regardless of what software they have on their desktop.
What services are available out there, and how much do they cost?
These are some of the most popular solutions available that are accessible via API and run in the cloud (absolute ease-of-use).
TauYou is a Spanish company located in Barcelona that provides machine translation via API as well as custom problem solving services for major localization companies. TauYou also provides natural language processing, which is great if you want to extract keywords, recognize topics, or categorize texts automatically.
Microsoft Hub, also known as Microsoft Translator Hub, is the method by which you can train your own private Microsoft Translator. It comes bundled with Microsoft Translator if you wish to use it.
A pretty good comparison page that compares Microsoft Translator vs.Google Translate is already on the web, although more such posts are sorely needed.
A video is also available.
Google Translate is a good option for many, but unfortunately, training does not seem to be available on Google Translate as it is with Microsoft Hub. Google takes their own actions to improve their quality, but without training, specificly preferred translations of words or sentences isn’t possible, so it won’t learn your glossary or style guide.
Asia online can do pretty much everything in terms of connecting to your system via API, training, and language pairs.
SAIC Omnifluent Translate
Omnifluent is a particularly attractive option for enterprises that are looking not only to machine translate text, but also to machine translate voice using voice recognition, or documents using their proprietary OCR.
Kantan MT is an intriguing option. They are launching in June, and offer cloud-based MT and easy training of your own system.
Bigger players like Google and Microsoft are able to swing deals with institutions and enterprises to train their machine translation systems using huge imports of data. Players like Kantan MT may have a harder time getting the data necessary to keep up with larger players like Microsoft or Google, but as you can train your own system, it could work out for cutting edge players.
Article written by the Wordbee S.A. team
Including linguistic technology (CAT tool, machine translation, advanced algorithms) and project management technology, Wordbee runs in your browser and allows everyone to collaborate in a shared tool/workspace. It’s the premier translation management system for the discerning translation agency or enterprise, and is fast becoming the preferred enterprise solution worldwide.