Neural Machine Translation Based on Multi Translation Parallel Corpus

In the context of continuous globalization, people’s demand for machine translation is also increasing. Neural Machine Translation (NMT) is a research hotspot in the field of machine translation. In the research of neural machine translation, traditional rule-based methods ignore the syntactic information between the source and target languages, resulting in poor performance of neural machine translation. This article adopted the Transformer model based on parallel corpora, combined with attention mechanism and syntactic information, to design a neural machine translation model. This article studied the construction of a multi translation parallel corpus, the design and development of a machine translation system, and the application of a multi translation parallel corpus in neural machine translation. The experimental results showed that on the same dataset, this model outperformed the traditional Sequence to Sequence (Seq2Seq) model. On different languages, the performance of this model was better than that of the Seq2Seq model. The Transformer model based on a multi translation parallel corpus can improve translation accuracy and coherence, and can achieve 89.6% accuracy when dealing with long compound sentence patterns.