Sentence Level Rhetoric Relations Extractor for Russian

Information on discourse structure of a text may be useful for some tasks of natural language processing and text generation, so the problem of analysis of discourse is one of the forward-looking areas of research in the field of computational linguistics. Rhetorical Structure Theory is a framework of analysis of discourse structure, which allows to describe the relationship between clauses in the text and represent it as a hierarchical tree-like structure. The implementation of automatic discourse analysis algorithm includes two subtasks: discourse segmentation into elementary units and discourse parsing. There are several automatic discourse parsers for English, and the first subtask can at the moment be performed on a state-of-art level, but not the second. And there is no discourse parser available for Russian yet. In the present work I will not consider the segmentation task and will instead center on intra-sentential parsing of discourse structure of Russian written texts. Thus the aim of this work is to create an algorithm of automatic discourse parsing for Russian sentences given that they were divided into elementary discourse units using supervised machine learning on an annotated corpus of Russian texts.