On the use of positive definite symmetric kernels for summary extraction

The task of creating a short, accurate and fluent summary starting from a larger text document or group of documents is called text summarization. When the summary is generated by extracting parts from the original text, the process is known as extractive summarization. This work is focused on the use of convex optimization and positive defined symmetric kernels for the extractive summarization of a text. The paper includes two new contributions. First, we show how the kernel method can be used for unsupervised extractive summarization. Second, we investigate empirically the performance of different kernel functions with respect to quality and execution time.