SMT is based on the comparison of parallel texts with high content volume and the calculation of the most probable translation. SMT has a feature of self-training, and translation quality with this approach depends directly on the volume of the parallel data for training. SMT engines today are mostly created on the basis of Moses, a free statistical machine translation engine.

Following the industry trend, PROMT researched Moses and developed its own approach for training an SMT engine on given parallel corpora.

Key Components:

  • Translation Model. Basically, this is a table with the following columns: source n-gram, target n-gram, and probability that this source n-gram translated with this target n-gram. The Translation Model is used to create translation candidates.
  • Language Model (LM). The LM is a statistical model of target language with probabilistic characteristics of n-grams in target language. It is built on target corpora and used for evaluation translation candidates created by the Translation Model.

Key Advantages:

  • Fast and fully automated engine training (in most cases, language-independent process)
  • Fluency (since the MT output is built from pieces of human translated texts)

As an SMT engine is trained on given texts, PROMT trains a client-specific SMT engine for a dedicated client. PROMT uses its own algorithms to correct the alignment in the client's corpora and to take into account the specifics of the target language.

Contact PROMT

Ask your question!


PROMT is the world leading provider of automated translation software. Our applications and technologies receive awards at major industry conferences and from the leading computer magazines. PROMT offers translation solutions for home and business use, as well as for corporate intranets and online projects.

PROMT provides machine translation for 20+ languages: English, Russian, German, French, Spanish, Italian, Portuguese, Chinese in two variants, Ukrainian, Kazakh, Turkish, Bulgarian, Latvian, Japanese, Polish and others, with a total of 80+ language pairs.

The Ukrainian, Latvian and Polish language pairs included in the PROMT solutions are developed by Trident Software.

Ask you question