Machine translation (sometimes called automated translation) is not to be used in every situation but, when applied correctly, there are several reasons why it is such a useful tool. Benefits include:
…all in a secure environment.
With Capita TI, our customers are assured a personal service, enabling bespoke engine customization. Our wealth of experience and range of machine translation (MT) services enable us to offer two key machine translation solutions that address an extensive array of applications:
Benefit from our MT technology and software with complete assurance of data security through a locked-down hosted MT solution, or on-site machine translation installation.
Seamlessly apply machine translation to your standard localization workflow to reduce costs and improve the productivity and consistency in your translations.
Machine translation is not suitable for every translation requirement. Before we undertake any work for any of our customers, we always conduct a pilot project. Let us assess your existing localization model and requirements in order to calculate ROI before blindly investing in an MT solution.
As part of our MT solutions, we also offer a range of MT services including the collection of data, engine customization, and MT performance evaluation. You can read more about our MT services here.
Automated Translation can be applied in different ways to help solve a number of localization challenges, but there are areas where it might not add value – for example, we would never recommend solely using an MT engine for the translation of marketing material.
We believe that MT is more than simply applying technology. A full machine translation service is about understanding the complete content lifecycle and evaluating how MT can be introduced into localization workflows to solve real business challenges.
We offer a consultation service to evaluate your localization processes and requirements, recommending suitable solutions through the application of our MT technology. If you are evaluating MT within your localization process we are willing to engage with you from the outset and offer an open book analysis of a project workflow, sharing all knowledge in order for you to make an informed decision.
Data is the fuel to building high quality, customized machine translation engines, so collecting, preparing, and cleaning data is key. Working in close collaboration with you, our goal is to maximize the use of content-specific translation memories (TMs) and glossaries in order to improve performance. All data will undergo extensive cleansing and normalization before being used to build the relevant MT engines.
Over the years our team has built up an extensive amount of openly available corpora, organized by industry domain. industry domain. For customers that don’t have a lot of their own data, our corpus acquisition service can combine any available customer data with relevant datasets that are at our disposal, to enable our customers to get on the customized MT ladder.
Our MT building and tuning strategy is multi-phased and based on continual development. We begin the engine-building process by profiling the content for our customers’ localization needs to the correct engine type. Corpora, including TM and glossaries, are added to this model to build the engine. Using a combination of automated and human testing, we then assess the output of that engine from both a quality and productivity perspective and are then able to tune the engine to further enhance its capabilities.
Building an engine is only the start of your MT journey, but what benefit are you really getting from using MT – and how can these benefits grow? Capita TI offers an on-going engine performance evaluation and reporting functionality.
One very important aspect of creating an MT environment is ensuring that there is a standard way of evaluating the output from each engine and making sure that feedback is incorporated into the engine tuning process. As well as scoring the engines automatically using industry metrics, e.g. BLEU and TER (Translation Edit Rate), we also benefit from our experienced network of qualified translators to gather useful feedback. This feedback is used by our MT engineers and linguistic computational experts to make any necessary tweaks to the engines to ensure that they produce translations that are more focused on the relevant content type.
Post-edited machine translation (PEMT) is the examination and correction of text generated from an automatic machine system. It is the process of improving a machine-translated text, to produce either an understandable text (Light PEMT) or a publication quality translation (Full PEMT) that would be comparable to a high quality human translation.
Capita TI offers both levels of PEMT. Light PEMT involves performing as few modifications as possible on the text to make it a semi-accurate reflection of the source text content, with less consideration of writing style and fluency. Full PEMT involves all standard revision practices as well as making syntactic/grammatical modifications to adapt the deviant linguistic nature of some MT output to standard stylistic practice in the target language.