The Chinese Discourse Parser

Discourse relation analysis usually focuses on identifying elementary discourse units (EDUs) and recognizing the discourse relation between EDUs. Previous work on this topic primarily focused on English. To the best of our knowledge, there is no publicly available tool for Chinese discourse relation analysis. In the work, based on the human-annotated corpus, the first end-to-end discourse relation analysis system for Chinese, which includes elementary unit segmentation, discourse connective disambiguation, explicit relation recognition, and implicit relation recognition has been developed. In the experiments, the model achieves a F-score of 89. 8% in explicit relation recognition and 55. 5% in implicit relation recognition.