作者: Slav Petrov , Dan Klein
DOI:
关键词: Computer science 、 Treebank 、 Context-free grammar 、 Grammar 、 Top-down parsing 、 Parsing 、 S-attributed grammar 、 Rule-based machine translation 、 Parser combinator 、 Generative grammar 、 Natural language processing 、 Artificial intelligence 、 Parsing expression grammar 、 L-attributed grammar
摘要: Treebank parsing can be seen as the search for an optimally refined grammar consistent with a coarse training treebank. We describe method in which minimal is hierarchically using EM to give accurate, compact grammars. The resulting grammars are extremely compared other high-performance parsers, yet parser gives best published accuracies on several languages, well generative numbers English. In addition, we associated coarse-to-fine inference scheme vastly improves time no loss test set accuracy.