Statistical Machine Translation

Statistical machine translation (SMT) is a machine translation paradigm that generates translations based on a probabilistic model of translation process, the parameters of which are estimated from parallel text. Modeling, training, and decoding are three fundamental problems in SMT. SMT has evolved from an early word-based approach to recent phrase-based and syntax-based approaches in the past decades. As a data-driven technology, it will continue to develop in the era of big data.