An Analysis Of Compositions By Chinese Students And Native Speakers Based On Rhetorical Structure Theory

By annotating Chinese students' and native speakers' compositions of the same topic based on RST, this thesis describes and compares the rhetorical structure diagrams of these compositions from the perspective of the amount, frequency and distribution of each relation, especially the frequency and position of the joint relation and the top structure of their diagrams to reveal the differences between them and shed light on the deficiencies of Chinese students' compositions.It is a new research perspective to analyze Chinese students' compositions based on RST. By revealing the discourse relations at both hierarchical and linear levels of a text, RST provides a clear and objective way to see the manner of writing and the logic of sentences. It helps scorers or teachers see students' thinking pattern more clearly and know the deficiencies of Chinese students' compositions better, hence providing reference value for building a better scale for grading compositions.This thesis not only recounts our experience in RST Treebank building, but also provides our findings of the study by examining three hypotheses. First, the RST diagrams of Chinese students' compositions bear a lower variety of discourse relations, but they share some same discourse relation preferences with the diagrams of native speakers' compositions. Second, the diagrams of Chinese students' compositions bear more joints and joints occupy different positions in the diagrams of Chinese students' and native speakers' compositions. Third, the top two levels of the diagrams of compositions by Chinese students and native speakers are mainly the same and show same discourse relation preferences in the top two levels.Our study shows that the diagrams of compositions by Chinese students and native speakers share more similarities than differences. In particular, they have similar top structures. The marked difference lies in a higher variety of relations and a lower frequency of joint found in the diagrams of compositions by the native speakers studied.