TJU GSummary at TAC2011: Category oriented extractive content selection for guided summarization

The guided summarization aims to capture evolving information of a single topic changing over time under the background of much emergency happenings. It delivers salient and novel information to a user who has already read a set of older documents covering the same topic. Topic is category oriented, however, there is no query description. Therefore, guided summarization raises new challenges. In this paper, the category oriented extractive content selection method for guided summarization is proposed, which is completely language independence. Meanwhile, we submitted two systems. Our systems rank top 5 under PYRAMID metrics among 48 running systems. However, we rank middle under ROUGE and BE averagely. Our methods show the great difference for different evaluation metrics. Therefore, we try to think why this happens? What is the best method? What is the best evaluation metrics?