Abstract

Summary assessment is a complex task, due to the problems related to sistematicity and to the time it takes. These problems have motivated the study of reliable automatized assessment methods. In this scenario, the aim of this work is to identify an efficient method for assessing summaries, based on Latent Semantic Analysis (LSA). The summaries were written by students from secondary school in the city of Valparaíso, Chile. To acomplish the mentioned aim, we correlated the scores asigned by three teachers to 244 summaries of predominantly expository texts and 129 summaries of predominantly narrative texts with the scores provided by three computational methods, based on LSA. The methods are: 1) comparison of summaries with the source text, 2) comparison of summaries with a summary developed by the consensus of a group of linguists, and 3) comparison of summaries with three summaries constructed by three language teachers. General results show that the scores asigned by method 2 and 3 are statistically similar to the scores asigned by the teachers, when the source texts are predominantly narrative. However, this similarity is not so when the source text is predominantly expositive. © 2011 PUCV, Chile.