Probability Distribution of Discourse Relations Based on a Chinese RST-annotated Corpus

Abstract

This article investigates the probability distribution of discourse relations with respect to the categories of the rhetorical structure theory (RST). Twenty texts were randomly extracted from a Chinese RST-style discourse corpus, and the frequency distribution of different discourse relations was investigated. Results show that all the RST relations in all the 20 texts fit perfectly well with a modified right-truncated Zipf-Alekseev distribution. The investigation verifies that at discourse level, linguistic development also follows the diversification process.

5. Acknowledgements

We thank JQL referees for insightful comments. This work is partly supported by the National Social Science Foundation of China (Grant No. 08CYY027 and 09BYY024).

Notes

¹O'Donnell, M., retrievable on May 20, 2010 from http://www.wagsoft.com/RSTTool/index.html