中文摘要：

图解码依存分析方法是一种重要的依存分析方法,优点是解码具有全局最优的特点,能够找到模型意义下的全局最佳依存树。传统图解码依存分析模型大多采用基于特征的线性评分模型,常常需要选取大量的人工特征,这一方面耗时费力,加剧了模型过拟合的风险,另一方面也显著降低了系统的运行效率。同时由于采用子图分解策略,传统图解码分析中的特征提取严重受到子图规模的限制,无法提取具有全局意义的分析特征。深度图解码依存分析研究部分解决了这些问题,本文概要介绍了近年来几个代表性的深度图解码依存分析研究工作,总结了国内外在深度图解码依存分析方面的现状和进展。

英文摘要：

Graph-based approach is one of the most successful approaches to dependency parsing. It is at- tractive for capability of offering global inference over space of all possible trees, and thus guarantees to find the best-scored trees given a tree scoring model. Traditional graph-based dependency parsing models usually adopt linear feature-based scoring models, which heavily rely on time-consuming feature engineer- ing. The huge number of features they involves also dramatically slows down the parsing speed. Typical graph-based models factor the dependency tree into subgraphs, which limits the scope of feature extraction to the subgraph and inhibits the performance of recovering long distance dependencies. Recent work intro- duces deep learning models into graph-based dependency parsing models and seems to partially solve or al- leviate the problems. In the paper I survey some of this work and present the advances they have achieved.