Uyghur Language Model with Graphic Structure

This paper describes a novel agglutinative language modeling strategy for Uyghur with graphic language model as structure. In graphic modeling language model, sentences are organized by morphemes as a directed graph, which is different from the linear structure in n-gram language models. The graphic language model is verified in two typical natural language processing application scenarios, morphological analysis and machine translation. The experiments show that the graphic language model achieves significant improvement in both morphological analysis and machine translation.