“…Most previous work on automatic analysis of information structure relies on supervised learning (Teufel and Moens, 2002;Burstein et al, 2003;Mizuta et al, 2006;Shatkay et al, 2008;Guo et al, 2010;Liakata et al, 2012;Markert et al, 2012). Given the prohibitive cost of manual annotation, unsupervised and minimally supervised techniques such as clustering (Kiela et al, 2014) and topic modeling (Varga et al, 2012;O Séaghdha and Teufel, 2014) are highly important. However, the performance of such approaches shows a large room for improvement.…”