Keyword Extraction From Chinese Text Based On Multidimensional Weighted Features

This paper proposed to solve the problems of incomplete coverage and low accuracy in keyword extraction of Chinese text based on intrinsic feature of the Chinese language and an extraction method of multidimensional information weighted eigenvalues. This method combined theoretical analysis and experimental calculation to study the parts of speech, word position, word length, semantic similarity and word co-occurrence frequency in Chinese texts. By combining multidimensional data related to word frequency, word feature values, word similarity and word co-occurrence probability, we calculated that the weighted eigenvalues obtained by comparing precision rate, recall rate and F measure and concluded that the proposed method can give a better measure of the word accuracy than using word frequency or the basic eigenvalue methods alone. The conclusions obtained in this study provide reference values for keyword extraction and text mining. Subject Categories and Descriptors I.2.7 [Artificial intelligence]: Natural Language Processing Text analysis; H.2.8 [Database Applications]: Data mining; General Terms: Chinese text mining, sentiment analysis