Depending on content type, domain and language direction, MT quality may vary quite a lot. Clients can choose and/or evaluate available MT engines and deploy the best performing engine in a consistent and unified way. To guarantee continuity, a fallback option can be provided in case an engine or server goes down.
Our strategy is to implement MT as an integrated process step within enterprise applications, including Content Management Systems (CMSes), support systems and localisation workflows. The CrossLang Gateway serves as ‘software glue’ for the various interfaces being used in the translation process. Only one client-side interface implementation is required to communicate with all MT systems connected to the CrossLang Gateway server.
For real-time translation, enterprise applications can be seamlessly integrated with MT so that automatically generated translations are published immediately. If the automated translation output requires post-editing by human resources, a connection with a translation editing interface can be provided so that the final output meets the expected quality levels. In this case, obtaining time and cost savings as compared to the traditional translation approach is the main objective.
In addition to the more integrated approach, offline documents can be processed via a web portal. Filter support is available for MS Word, Excel, PowerPoint, HTML, XML and XLIFF.
Rule-based engines typically allow for lexical customisation, while statistical MT engines (such as Moses MT) are trained with a parallel corpus, resulting in the drawback of terminological variation and inconsistency in the MT output. A proprietary term protection routine can be activated in the CrossLang Gateway ensuring that company-specific terminology overrules the potentially more generic translation of source terms.
Most MT engines struggle with the positioning of inline tags in the target language. The tag handling routine in the CrossLang Gateway makes an educated guess concerning the appropriate position of tags in the target segment, greatly reducing the amount of required editing.
The CrossLang Gateway allows users to take the highly academic Moses SMT system out of the lab and put it at the heart of a real-life translation and localisation production environment, including full support for tag handling, named entity recognition, real casing and other text engineering capabilities.
The CrossLang Gateway enables an efficient production deployment of custom built open-source Moses SMT engines, ensuring redundancy and load balancing.
Additional pre/post processing steps that are required to provide a fully customised solution rather than just a default Moses system can be designed and implemented on request.