A Study On The System Of Linguistic Features In Contemporary Chinese Text

In order to questing for a more scientific model of examinee evaluating, Tsinghua University sponsored the item of "the System of Evaluating and Selecting the Persons with Bility that Based on the Internet Environment" (viz. "the System of Internet Evaluating"). For improving the reliability and validity, "the System of Style-identification on Internet" is specially devised.This study is based on the theories of Corpus Linguistics, Statistical Linguistics, Algebraic Linguistcs and Style Linguistcs, and used the methods of induction, deduction, antithesis, etc. The object is trying to choose the salient linguistic features that can discriminate the different authors, and then some linguistic model would be set up.The "MF/MD" means "multi-dimensional"and "multi-feature". Biber set up this model to study the differences between Speech and Writing British English. The MF/MD approach is based on factor analysis. In this paper we will undertake a genre analysis of different authors' linguistic features.The paper used factor analysis to reduce 45 linguistic features to 6 dimensions or factors. The identified 6 dimensions or factors include: ascensive mood versus wateriness mood; miscellaneous expressing versus inornate expressing; narrative versus involved; orderly inclination versus inorderly inclination; Speech inclination versus Writing inclination; modificatory and descriptive versus complementary and explanatory. Every dimension provides a good review of the functions of each of these linguistic features. Factor analysis is the primary statistical tool of the MD/MF approach to textual variation. In this study, the frequencies of linguistic features, are reduced to a small set of derived variables, the "dimension" or "factor". Each factor represents an area of high shared variance in the original data, a group of linguistic features that can co-occur with a frequency. As these factors underlie linguistic features, they are conceptually clearer than the many features considered individually. Throught the same method the validating process can be done. Whether the appointed text attribute to the author or not is decided by its dimension score of a genre, that can be obtained by adding together the mean factor scores of all features with positive weights on a factor and then subtracting the mean factor scores of all features with negative weights on the same factor: The MF/MD approach is useful and significant. It is found that this approach, while providing a powerful tool for genre analysis, is quite demanding in time , labour and expertise in statistics.