This paper addresses the problem of creating a summary by extracting a set of sentences that are likely to represent the content of a document. A small scale experiment is conducted leading to the compilation of an evaluation corpus for the Greek language. Two models of sentence extraction are then described, along the lines of shallow linguistic analysis, feature combination and machine learning. Both models are based on term extraction and statistical filtering. After extracting the individual features of the text, we apply them to two neural networks that classify each sentence depending on its feature vector, the term weight being the feature with the best discriminant capacity. A three-layer feedforward network trained with the highly popular backpropagation algorithm and a competitive learning self-organizing map characterized by the formation of a topographic map, both trained on a small manually annotated corpus of summaries, perform the sentence extraction task. Both methods could be used for rapid light information retrieval-oriented summarization.