[目的/意义]文章旨在探究将不同语义知识融入机器翻译模型能否增强机器翻译的效果以及何种语义知 识的作用更为显著，以助力机器翻译研究与中华优秀传统文化的传承与传播。[方法/过程]研究选取了30万对精 加工的《二十四史》“古代汉语-现代汉语”平行语料作为实验数据，基于神经机器翻译OpenNMT模型，通过三种不 同的特征融合方法，将词边界知识、词性知识、实体知识和依存句法知识分别融入机器翻译模型的训练过程中。 [结果/结论]不同语义知识与模型的融合对典籍翻译效果有不同的影响，词边界知识、词性知识、实体知识对机器 翻译任务有一定的贡献且实体知识的贡献最大，依存句法知识无明显作用。

[Purpose/significance] This article aims to explore whether integrating different semantic knowledge into machine translation models can enhance the effectiveness of machine translation and which type of semantic knowl⁃ edge plays a more significant role. The purpose is to support the research in machine translation and the inheritance and dissemination of Chinese excellent traditional culture. [Method/process] The study selected 300,000 pairs of me⁃ ticulously processed "Ancient Chinese-Modern Chinese" parallel corpora from the "Twenty-Four Histories" as experi⁃ mental data. Based on the neural machine translation model OpenNMT, it integrated word boundary knowledge, partof- speech knowledge, entity knowledge, and dependency syntax knowledge into the training process of the machine translation model through three different feature fusion methods. [Result/conclusion] The integration of different se⁃ mantic knowledge with the model has varying impacts on the translation effectiveness of classical texts. Word boundary knowledge, part- of- speech knowledge, and entity knowledge contribute to the machine translation task, with entity knowledge making the largest contribution, while the role of dependency syntax knowledge has no obvious effect.