Semantic MMT Model Based on Hierarchical Network of Concepts in Chinese-English MT.

To study the generation of the semantic tree of Chinese sentence in Chinese-English Machine translation (MT), a new semantic-analysis model of Chinese multiple-branched and multiple-labeled tree (MMT) based on the hierarchical network of concepts (HNC) is proposed. Supported by word and rule knowledge-base of HNC, the model executed the semantic analysis using static and dynamic labels as a complex feature of MMT instead of a single feature of phrase structure grammar, and generated a HNC-MMT semantic tree for deep understanding of the semantic of Chinese sentence. Based on the semantic tree generated, the model can realize structure conversion of semantic chunks, and utilize a hybrid strategy of the statistical and rule-based to translate. Experiment shows one of the most important tasks of semantic analysis of Chinese sentence, the global eigen-chunk recognition, achieves accuracy above 85%, verifying the effectiveness. The model has been applied to system development of MT based on HNC. © 2013 ACADEMY PUBLISHER.