THUMT-Theano 3 : the original project developed with Theano 4, which is no longer updated because MLA put an end to Theano. The training criterion is maximum likelihood estimation (MLE). 2 It implements the sequence-to-sequence (Seq2Seq) (Sutskever et al., 2014), the standard attention-based model (RNNsearch) (Bahdanau et al., 2015), and the Transformer model (Transformer) (Vaswani et al., 2017). THUMT-TesnorFlow 1 : a new implementation developed with Tensor- Flow. It currently has two main implementations: 1. THUMT is an open-source toolkit for neural machine translation developed by the Tsinghua Natural Language Processing Group. Capable of learning representations from data, NMT has quickly replaced conventional statistical machine translation (SMT) (Brown et al., 1993 Koehn et al., 2003 Chiang, 2005) to become the new de facto method in practical MT systems (Wu et al., 2016).

Recent several years have witnessed the rapid development of end-to- end neural machine translation (NMT) (Sutskever et al., 2014 Bahdanau et al., 2015 Vaswani et al., 2017). With the availability of bilingual, machine-readable texts, data-driven approaches to machine translation have gained wide popularity since 1990s. 1 THUMT: An Open-Source Toolkit for Neural Machine Translation The Tsinghua Natural Language Processing Group January 1, Introduction Machine translation, which investigates the use of computer to translate human languages automatically, is an important task in natural language processing and artificial intelligence.

Windows toolkit 2 5 3