Custom, AI boosted engines with domain adaptation
Globalese engines can be adopted to the content, style and terminology of your existing legacy translations. You can simply choose the relevant training data (typically a Translation Memory), assign it to the engine and start training. Everything else is done automatically by Globalese.
You can train the following engine types:
- Stock+ engines
- Domain Adapted Engines
- AI-boosted Domain Adapted engines
Technically speaking, Globalese is language-independent. If you have the right amount of training data, you can practically train an engine for any language combination.
Fortunately, there is no need to have a large training dataset on your own. It is even better to use a smaller but more qualified training dataset instead of a large, unqualified one. We provide stock corpora for over 130 language combinations that can be used for standard or AI-boosted Domain Adapted engines. You can check the supported languages for Stock Corpora here.
Additionally, we provide Stock Engines that can be used as a basis for Stock+ custom engines for over 80 language combinations. You can check the supported languages for Stock Engines here.
Translation tool integration
You can easily integrate your custom Globalese engines into your translation workflow. We support most of the common CAT tools. There are two main types of integration: connectors and plugins.
Connectors are typically used to connect your Globalese system with your CAT tool server to retrieve and update training data.
Plugins are used to run translation projects. There are two types of plugins available:
Cloud Text Translation: you can use the Cloud Text Translation plugins either for pre-translating your projects or to get online propagation during translation. This is recommended for any standard scenarios.
Document Translation: you can use the Document Translation plugins for batch-pretranslation of projects. This is recommended for large projects.
We know how important process and workflow automation is, which is why we are offering API integration as a standard offering without any additional fee. You can use the API for almost all scenarios, such as
- uploading or updating training data,
- triggering engine training,
- deploying trained engines,
- running batch or real–time translations.
Advanced tag handling
Handling formatting and placeholder tags can be a real challenge. We are proud to say that Globalese has an advanced tag-handling feature, which ensures that tags are handled technically properly and placed correctly back into the translation in most cases. This is especially important when using machine translation for complex, XML-based, heavily tagged content, such as technical documentation or web shop catalogs.
We know how important security and confidentiality are. Therefore, Globalese is providing complex options for how you can fine–tune the access rights of your users to the different features and content in the system.
As an example, you can define Training Data Manager, Engine Manager, and Project Manager rights on the user and group level. Groups can be freely defined depending on your organization. For example, a group can represent a department, a product line, or (if you are an LSP) your customers. This way, you can keep the training data separate and avoid unintended access or usage in engines.
Cloud or on-premise
You can use Globalese as a cloud-based SaaS solution or as an on-premises installation. Please note that, in the case of an on-premises system, there are some restrictions on features due to technical reasons. Additionally, the exact configuration and price tag can vary greatly depending on the expected performance and availability. Please contact us for further details.
On our roadmap
We have following main new features on our roadmap:
With this feature, users can add keyword lists to an engine which are prioritized during translation. Even better, our plan is that the lists can be changed dynamically, without the need to re-train an engine. The feature will be available for the AI-boosted engines. Planned availability is in Q4 2123.
With this feature, users can define custom prompts on engine level, influencing the MT output. The feature will be available for the AI-boosted engines. Planned availability is in Q4 2123.