IBM recently rolled out the beta version of Model Customization for the Watson Language Translation Service. The new capability allows near real-time language translation based on models for particular requirements, intended for use in situations such as multilingual content curation and live chat support.
The new capability works by deriving custom models from base models specified by users, allowing them to pile custom facets to a base model to suit particular needs. For instance, when monitoring brand mentions, users can specify a base language pair to work with, and then add custom aspects such as ensuring the brand slogan is translated correctly.
The IBM Watson Language Translation Service itself has only recently reached General Availability, and the beta version of Model Customization is one of many steps that IBM is taking to add more power to the service, and the company intends to provide more APIs for user customization and tooling for simplification. As these are rolled out, more complex customization of models can be implemented.
The beta capability is expected to bring more value to the fledgling service that is already putting great effort into language recognition, bi-directional translation of published content, and models such as conversations and patents.
The Language Translation Service works with language pairs involving English, French, Spanish, Portugese, Arabic, Chinese, and Korean, depending on the domain (published content, patents, or conversational models). It can identify a far longer list of languages in plain text format.