The idea of machine translation dates back to 1629. However, machine translation really developed after 1954. By the end of the 1980s, many new methods of machine translation were born. Especially the statistical translation method (SMT) and the example-based translation method (EBMT). These two methods began to replace the rules of syntax and semantics by processing large blocks of text, which computing power allowed.
Machine translation according to statistical models has proved to be an effective method for developing high quality machine translation systems as SMT translation machines have consistently prevailed in annual machine translation competitions. Methods to increase the quality of the basic SMT translation system are still being of interest to many researchers with the aim of obtaining treatments that match the characteristics of specific language pairs such as word order correlation, choose words that match the context with the syntax processing methods in the pre- and post-processing steps. Realizing many advantages, we have decided to build a translation machine system using the SMT method as the foundation and hybridize SMT, EBMT, RBMT together in the hope of making a difference in the quality of the translation machine.
The translation machine system is a distributed system, the translation machines are scattered in the network environment (LAN / Internet), connected together by a load balancer and buffer for the purpose of increasing capacity. function to serve many concurrent users but still ensure the quality of automatic translation. Currently, we have built translators for Law and Medical topics, in which the translation quality of the translator on Law subject has reached a high level of usefulness. The translation machine for other topics is continuing to be built; We will publish translation engines on these topics whenever the translation quality meets the quality standards set by VIEGRID.