Abstract

We evaluate a novel method for automatic text summarization through text extraction. It attempts to find the summary most similar to the original text, thus giving an overview of all the contents. It also evaluates whole summaries, making no judgments on for instance individual sentences. A greedy search strategy is used to search through the space of possible summaries and select the best summary of those found. When evaluated on English abstracts from the Document Understanding Conferences our method performed fairly well. In this paper we evaluate it on Swedish human produced extracts. It performs poorly, which was expected since these extracts were not produced to reflect the whole contents of the texts. They only cover the most important topic.