Evaluation Metrics for the Summarization Task

Evaluation Metrics for the Summarization Task

Paulo Cesar Fernandes de Oliveira (Anhanguera University, Brazil)
DOI: 10.4018/978-1-4666-5019-0.ch005


Summary evaluation is a challenging issue. It is subjective, costly, time consuming, and, if it is human-assisted, can generate some bias. Due to this, several attempts have been made in the last decades in order to avoid all these drawbacks. Those attempts focused on automatic summary evaluation. This chapter provides a comprehensive survey about studies that have dealt with this topic and proposes and describes the development of an automatic summary evaluation method. In addition, it presents some experiments that have been carried out in order to assess the method’s performance.
Chapter Preview


Due to a crescent interest in summary evaluation in the past several years, some studies will be described in this section. In order to do this description in a more systematic fashion, we will follow an interesting classification proposed by Tucker (1999). He suggested four approaches, namely: direct evaluation; target-based evaluation; task-based evaluation and automatic evaluation.

Complete Chapter List

Search this Book: