This article reports an experiment with a computational simulation of an Information Retrieval System constituted of a textual indexing base from a sample of documents, an artificial neural network software implementing Adaptive Resonance Theory concepts for the process of ordering and presenting outputs, and a human user interacting with the system in query processing. The goal of the experiment was to demonstrate (i) the usefulness of Carpenter and Grossberg (1988) neural networks based on that theory, and (ii) the power of semantic resolution based on sintagmatic indexing of the SiRILiCO approach proposed by Gottschalg-Duque (2005), for whom a noun phrase or proposition is a linguistic unity constituted of meaning larger than a word meaning and smaller than a story telling or a theory meaning. The experiment demonstrated the effectiveness and efficiency of an Information Retrieval System joining together those resources, and the conclusion is that such computational environment will be capable of dynamic and on-line clustering with continuing inputs and learning in a non-supervised fashion, without batch training needs (off-line), to answer user queries in computer networks with promising performance.
Information retrieval system; Noun phrase; Semantics; Syntagmatic indexing; Text mining; Artificial neural networks; Adaptive resonance theory; ART neural networks; Computational simulation; Artificial intelligence
