The Research Of Computer Assistant Analysis On Chinese Song Poems' Style

As a representative type of Chinese ancient literature, Song Dynasty Poetry has been studying by lots of scholars and researchers since Song Dynasty, many of who devoted themselves to the study on the style of Song Dynasty Poetry. And this kind of research mainly focuses on the area of literae humaniores. However, the major task of this paper is to do some research of computer assistant analysis on the style of Song Dynasty Poetry. The relevant work of this paper includes aspects mentioned as follows:First of all, this paper makes a summarization of computer analysis on the style of poetry. In this part, the paper introduces different kinds of main methods as well as their skeletons and implemental steps. After that, this paper aims to find out not only their differences but also their advantages and disadvantages in order to propose a suitable method for our research on the computer-assisting style classification.Secondly, build and improve the corpus of Song Dynasty Poetry. According to the rules and forms of Song Poems, sentences are divided into sub-pieces. Closely combined words are extracted by calculating the frequency, mutual information and collocation rate. Strategies such as conditional probability are used to implement gelǜ-tagging and part-of-speech-tagging.Furthermore, according to the traits of Song Poems' expression, this paper proposes a text representation model which is suitable for Song Poems' style classification. Based on Apriori algorithm- classical algorithm of data mining to find associating words, this paper finds out words that appear concurrently in one text. We use this method to save more information of Song Poems' style. What is more, this paper makes full use of How Net to make our algorithm a bit understand the meaning of words. The main contents used in How Net are called concepts. By combining words and frequent associating words, the paper maps from words and frequent associating words to concepts, and proposes VSM which is based on two methods mentioned above to present Song Poems. The following experiments prove that our VSM is effective.Finally, this paper introduces different methods of text classification and the main ideas of these methods. According to the requirement, the paper chooses one of them, that is KNN Algorithm, and use Genetic Algorithm to select weights of terms in the VSM. And then, the paper improves KNN Algorithm's parameter in order to make it fit for Song Poems' style classification. In the end, this paper implements a classification algorithm for different style of Song Poems, and gets satisfying results.