A new method for classifying and searching software components by using a self-organizing neural network architecture

The method presented in this paper aims to simplify the construction of software component repositories. The repository makes possible the reuse of components, reducing the software implementation costs. The proposed method extracts informations from component documentation, or either, terms which compound the metadata to represent components. The components are automatically grouped, using the terms, in the repository by means of the ART-2A self- organizing artificial neural network architecture. The vectorial search strategy is used to retrieve software components which are grouped by the neural network. Experiments showed that this strategy improved the ordinary vectorial search by an average of 9.55% in precision, maintaining a similar quality in recall. This method also presented an relevant increase in the search performance.