Keyword extracting using auto-associative neural networks

Abstract

The large amount of textual information digitally available today gives rise to the need for effective means of indexing, searching and retrieving this information. Keywords are used to describe briefly and precisely the contents of a textual document. In this paper we present a new algorithm for keyword extraction. Its main goal is to extract keywords from text documents written in Spanish quickly and without requiring a large training set. This goal was achieved using auto-associative neural networks, also known as <i>autoencoders</i>, trained using only the terms designated as keywords in the training set, so that these networks can learn the features characterizing the important terms in a document.