RST-style discourse parsing plays a vital role in many NLP tasks, revealing\nthe underlying semantic/pragmatic structure of potentially complex and diverse\ndocuments. Despite its importance, one of the most prevailing limitations in\nmodern day discourse parsing is the lack of large-scale datasets. To overcome\nthe data sparsity issue, distantly supervised approaches from tasks like\nsentiment analysis and summarization have been recently proposed. Here, we\nextend this line of research by exploiting distant supervision from topic\nsegmentation, which can arguably provide a strong and oftentimes complementary\nsignal for high-level discourse structures. Experiments on two human-annotated\ndiscourse treebanks confirm that our proposal generates accurate tree\nstructures on sentence and paragraph level, consistently outperforming previous\ndistantly supervised models on the sentence-to-document task and occasionally\nreaching even higher scores on the sentence-to-paragraph level.\n