关键词: NLP, k-means++聚类分析, 方差分析, 唐代诗歌, 特征分析

Abstract: Tang poetry is the cultural treasure of our country, with a large quantity, diverse styles andthemes. In order to discuss the characteristics of Tang poems of different schools, based on natural lan-guage word processing, this study analyzed the stylistic characteristics and differences of the poems of fivemajor schools , namely , landscape pastoral poetry , frontier poetry , romantic poetry, realistic poetry and epic poetry , using statistical methods such as k-means + + cluster analysis, repeated measurement variance a.nalysis and paired sample t test. The results show that Tang poetry itself has strong commonality, such as“* thousands of miles" , “ten thousands of miles" , “where" and other keywords 'TF-lDF values are relativelyhigh in different schools. In terms of poetry characteristics of different schools, romantie poetry and realisticpoetry , frontier poetry and epic poetry , romantic poetry and epic poetry have high similarities in the use ofkeywords , and the p-values of TF-DF values of different keywords after paired sample t-test are 0. 9730. 383 and 0. 052 , respectively. There are significant differences between the other genres.

Key words: NLP, k-means + + clustering analysis , variance analysis, the Tang Dynasty poetry, featureanalysis.