This article discusses a summarizer system of documents named SASI. This system features an innovative approach to provide automatic summaries, based on the determination of the maximum independent subset of vertices, modeling the problem a graph of phrases (vertices) and the relationships between them (edges). The concepts and operation of the proposed summarizer and a series of tests comparing the results provided by SASI with others summarizer systems are described. Initial results are promising, evaluating questions of informativeness of the produced summaries on the parameters of time and algorithmic complexity.