为基于真实语料进行句法分析,构建了大规模的短语结构树库和依存结构树库,并尝试在两种结构的树库之间进行转换.讨论了宾州中文树库(Penn Chinese Treebank,CTB)中短语结构树库和依存结构树库的关系,并基于现代中文依存文法制定了中心...为基于真实语料进行句法分析,构建了大规模的短语结构树库和依存结构树库,并尝试在两种结构的树库之间进行转换.讨论了宾州中文树库(Penn Chinese Treebank,CTB)中短语结构树库和依存结构树库的关系,并基于现代中文依存文法制定了中心子节点过滤表,依据该表将短语结构的CTB转换为依存结构树库.在CTB中随机抽取200句语料,转换正确率达到了99.50%.基于该转换得到的依存结构树库可以进一步进行中文依存关系解析的研究.展开更多
Part of Speech (POS) Tagging can be applied by several tools and several programming languages. This work focuses on the Natural Language Toolkit (NLTK) library in the Python environment and the gold standard corpora ...Part of Speech (POS) Tagging can be applied by several tools and several programming languages. This work focuses on the Natural Language Toolkit (NLTK) library in the Python environment and the gold standard corpora installable. The corpora and tagging methods are analyzed and com- pared by using the Python language. Different taggers are analyzed according to their tagging ac- curacies with data from three different corpora. In this study, we have analyzed Brown, Penn Treebank and NPS Chat corpuses. The taggers we have used for the analysis are;default tagger, regex tagger, n-gram taggers. We have applied all taggers to these three corpuses, resultantly we have shown that whereas Unigram tagger does the best tagging in all corpora, the combination of taggers does better if it is correctly ordered. Additionally, we have seen that NPS Chat Corpus gives different accuracy results than the other two corpuses.展开更多
文摘为基于真实语料进行句法分析,构建了大规模的短语结构树库和依存结构树库,并尝试在两种结构的树库之间进行转换.讨论了宾州中文树库(Penn Chinese Treebank,CTB)中短语结构树库和依存结构树库的关系,并基于现代中文依存文法制定了中心子节点过滤表,依据该表将短语结构的CTB转换为依存结构树库.在CTB中随机抽取200句语料,转换正确率达到了99.50%.基于该转换得到的依存结构树库可以进一步进行中文依存关系解析的研究.
文摘Part of Speech (POS) Tagging can be applied by several tools and several programming languages. This work focuses on the Natural Language Toolkit (NLTK) library in the Python environment and the gold standard corpora installable. The corpora and tagging methods are analyzed and com- pared by using the Python language. Different taggers are analyzed according to their tagging ac- curacies with data from three different corpora. In this study, we have analyzed Brown, Penn Treebank and NPS Chat corpuses. The taggers we have used for the analysis are;default tagger, regex tagger, n-gram taggers. We have applied all taggers to these three corpuses, resultantly we have shown that whereas Unigram tagger does the best tagging in all corpora, the combination of taggers does better if it is correctly ordered. Additionally, we have seen that NPS Chat Corpus gives different accuracy results than the other two corpuses.
