Machine Translation Engine


    Machine Translation Engine

    It is a machine translation engine based on artificial neural network, one that has evolved from Saltlux’s existing statistical machine translation engine. CAT (Computer-assisted translation) language resource management tool is integrated and configured so that machine translation learning can be specialized for each domain to improve translation quality and productivity.


    < Artificial neural network machine translation engine concept map >

    Main Features

    • Ease of domain application

    It is possible to perform specialized data processing in line with the data characteristics of customers who are adopting the machine translation engine, and to create domain-specific machine translation models. The generated model is combined with the base model trained by Saltlux’s machine translation engine to perform domain-specific machine translation as a new model.

    • Quality improvement through automatic translation post-processing and re-learning

    The processing structure of the machine translation engine consists of a virtuous cycle of NMT learning, translation application, assetization, and NMT relearning. With these systems, the more you use a translation engine, the higher quality and productivity you’ll expect from automatic translation.


    < Machine translation engine’s virtuous cycle work structure >

    • Connectivity to various office tools

    The machine translation engine provides the Office plug-in, which enables integration with office tools. You can use the translation results immediately within various office programs.

    Main features and specifications

    Machine translation engine consists of data area for searching and managing translation resources and various dictionaries, NMT area for performing and managing machine translation, and management tool area for system users to check and manage translation results.


    < Machine translation engine feature block diagram >

    •  Machine translation learning and model management

    It provides learning model management and evaluation features for quality control of machine translation. The learning model management function can generate a number of machine translation models by setting various parameters and changing the learning data, and support evaluation and comparison analysis functions for each generated model.

    • Learning data and dictionary management

    It provides learning translation data and translation dictionary management functions, so that you can apply machine translation to the domain, and better manage the quality of translation in that domain.

    •  Service management feature

    It provides various management features related to machine translation service such as API service management, machine user management, and machine translation service usage status search.

    • Translation result confirmation and post-translation editing management

    You can use this function to revise translation results. Changes in translation are reflected the learning model, affecting the next translation. Through this virtuous cycle of learning, translation, supervision and re-learning, the more you use the system, the better the quality of translation can be.

    Main engine screen

    Untitled web 14
    Untitled web 15