Neural Machine Translation System Based On Biomedical Corpus

The faster our society develop,the more communication among different languages.The emergence of machine translation has reduced the human effort for translation.However,no matter human translation or machine translation,the criteria of the accuracy and timeliness for translation are extremely high.Machine translation cuts labor costs in large number of translation tasks.Machine translation system can help people to perform translation work anytime and anywhere.From statistical machine translation to neural-network-based machine translation,both of them achieve satisfactory accuracy and efficiency.And they can replace human translation in some extent to meet the translation requirement in daily life.Based on the Transformer model,we add semantic disambiguation work and external dictionaries to build a translation model.The proposed translation model uses the sequence-to-sequence translation process which is the benchmark in machine translation based on neural networks.It abandons traditional recurrent neural networks,long and short-term memory networks,etc and is constructed with a Transformer model composed entirely of attention mechanism.In order to construct the specialized vocabulary of biology and medicine,we combined biomedical corpus which is obtained by a crawler system with general corpus.The experimental results show that based on the mixed corpus,the neural machine translation model not only ensures the accuracy of the translation of common-sense statement,but also performs more professional on the entire entities and sentences of biology and medicine.Based on the neural machine translation model for biomedical corpus,we constructed a web service.The translation service can help more researchers who are engaged in the field of biology and medicine to translate and understand foreign language literatures.