Large-Margin Learning of Submodular Summarization Models
Ruben Sipoš, Pannaga Shivaswamy, Thorsten Joachims · 2012
In this paper, we present a supervised learn-ing approach to training submodular scoring functions for extractive multi-document sum-marization. By taking a structured predicition approach, we provide a large-margin method that directly optimizes a convex relaxation of the desired performance measure. The learn-ing method applies to all submodular sum-marization methods, and we demonstrate its effectiveness for both pairwise as well as coverage-based scoring functions on multiple datasets. Compared to state-of-the-art func-tions that were tuned manually, our method significantly improves performance and en-ables high-fidelity models with numbers of pa-rameters well beyond what could reasonbly be tuned by hand. 1