Hierarchical machine translation method and system based on predicate argument structures

Abstract

The invention relates to a hierarchical machine translation method and system based on predicate argument structures. The hierarchical machine translation method comprises training and translating processes, and comprises the following steps: organically organizing all predicate argument structures in sentences into a graphic structure; taking the graphic structure as a top semantic framework structure for improving hierarchical phrase machine translation to directly build the models of the predicate argument structures in the hierarchical phrase machine translation, separately translate fragments with independent semantics and combine the fragments into a final translated text according to structures among the fragments. The translated text obtained by using the method is better in semantic similarity, semantic structure and long-distance ordering characteristics, thereby reducing the probability of generating meaningless translated texts through machine translation.

Description

technical field [0001] The present invention relates to the technical field of natural language processing, in particular to a machine translation method and system based on the semantic structure of predicates. Background technique [0002] For machine translation, semantic information is very important information. [0003] The invention patent "Semantics-Based Machine Translation System and Method" discloses a semantics-based natural language machine translation method, which is used to automatically translate the original text in the original language into translations in one or more target languages. The method includes the following steps : Extract a sentence of the original text; perform semantic analysis on the sentence according to the semantic unit representation library, so as to obtain the sentence meaning expression of the sentence; according to the semantic unit representation library, expand the sentence meaning expression with the semantic unit representation...

AI Technical Summary

Problems solved by technology

However, due to the relatively discrete organizational structure of the predicate argument structure, existing work uses this structure as supplementary information to guide translation rather than directly modeling it in machine translation

At present, the obstacles to directly modeling the predicate argument structure are: the predicate argument structure can be nested with each other, and the same word or phrase can be used as different argument components in different predicate argument structures; There may be words or phrases between the components of the predicate that do not belong to the predicate argument structure

Images