Formulation of Natural Language Processing based on Graph Theory

我们实现了一个可从部分标注语料库中训练的依赖解析器，其中只有少数单词标注了依赖信息。解析器与最先进的解析器一样精确。详细地说，解析器是基于最大生成树框架的。我们公开了自然语言处理工具。此外，我们还做了一个关于培训语料库的演示，并将其公开。

We realized a dependency parser trainable from partially annotated corpus in which only a few words are annotated with dependency information. The parser works as accurate as state-of-the-art parsers. In detail, the parser is based on a maximum spanning tree framework. We made the natural language processing tools publicly available. In addition, we made a presentation about the corpus for training and made it public.