Chart machine translation method and system based on rhetorical structure theory

The invention provides a chapter machine translation method and system based on a rhetorical structure theory. The method comprises the steps that original text information of a chapter original text is collected; the method comprises the following steps: segmenting original text information of a chapter original text into a plurality of sentences or clauses, wherein each sentence or clause serves as an EDU; the obtained EDU is analyzed, and an RST text structure tree is obtained; inputting the rhetorical relationship information of the RST chapter structure tree into a set rhetorical relationship matching rule to obtain a rhetorical relationship sequence of the chapter original text; inputting the rhetorical relation sequence into a chapter translation model, modeling the rhetorical relation sequence to obtain relation type embedding, and fusing the rhetorical relation sequence into the chapter translation model to obtain new word embedding; and embedding the obtained new words into an input encoder for trai