Method for solving generation of untranslated words based on medical neural machine translation

The invention discloses a method for solving generation of untranslated words based on medical neural machine translation. The method aims to translate a medical English abstract into a Chinese abstract. Therefore, a researcher feels convenient to read. The method comprises the following steps of: inputting the medical English abstract, preprocessing a text, calling a medical translation model, and judging whether the medical English abstract contains an untranslated word or not; and if an untranslated word appears in the translation result, calculating (unk) a source language position throughan ATTENTION mechanism, then using the word to carry out medical professional dictionary query, replacing (unk) in the translation result, and returning the translation result to the user. 本发明公开了一种基于医学神经机器翻译的解决产生未翻译单词的方法，该方法的目的是将医学英文摘要翻译为中文摘要，便于研究学者阅读，包括以下步骤：输入医学英文摘要、文本预处理、调用医学翻译模型、判断是否含有；如果翻译结果中出现未翻译单词则通过ATTENTION机制计算在源语言位置然后使用该词进行医学专业词典查询，替换翻译结果中的并返回给用户。