基于依存树的藏语语义分析

Dependency tree based Tibetan semantic dependency analysis

藏语语义依存分析是以藏语依存句法分析为基础的深层语义研究。该文从词法分析和句法分析等浅层研究出发,结合藏语自身语法结构和语义单位之间的关系特点,实现了藏语语义依存分析。在制定了藏语语义依存关系标注规范并设计了藏语语义依存关系特征模板的前提下,采用感知机进行了藏语语义依存分析模型的训练,经实验该模型在人工标注测试语料上的根准确率、依存弧准确率、依存弧类型准确率及完全准确率等4个指标分别达到了89.56%、78.63%、71.67%及32.32%,证实了该模型在藏语语义依存分析任务中具有良好的性能。

Tibetan semantic dependence analysis is a deep semantic study based on Tibetan-dependent syntactic analysis.This paper starts from the shallow research of lexical analysis and syntactic analysis,and combines the characteristics of Tibetan grammatical structure and semantic unit to realize the semantic dependence analysis of Tibetan for the first time.Under the premise of formulating the Tibetan semantic dependency labeling specification and designing the Tibetan semantic dependency feature template,the perceptual machine is used to train the Tibetan semantic dependence analysis model.The experimental results show that the root accuracy,dependency arc accuracy,dependent arc type accuracy and complete accuracy of the model on manual labeling test corpus reached 89.56%,78.63%,71.67%and 32.32%,respectively,which confirmed that the model has good performance in Tibetan semantic dependence analysis tasks.