Intelligent automatic generation of text summaries with Soft Computing techniques

Abstract

This thesis develops two different strategies to build automatic summaries of texts using Soft Computing techniques. The first uses a Particle Swarm Optimization technique that, from the vectorial representation of the texts, constructs an extractive summary combining adequately several punctuation metrics. The second strategy is related to the study of causality inspired with the management of uncertainty by the Fuzzy Logic. Here, the analysis of the texts is carried out through the construction of a graph by means of which the most important causal relationships are obtained together with the temporal restrictions that affect their interpretation. Both strategies fundamentally imply the classification of the information and reduce the volume of the text considering the recipient of the summary constructed in each case.