Abstract

PIEDRA DIEGUEZ, Laynier A.; TORRES LOPEZ, Surayne; MEDINA PAGOLA, José E.  and  RIVERO HECHAVARRIA, Claudia C.. Contextual cosine for semantic enrichment and comparison of short texts generated by Project Management tools.. Rev cuba cienc informat [online]. 2018, vol.12, suppl.1, pp. 151-162. ISSN 2227-1899.

The short texts analysis has been a problem to deal with, during the automatic decision support processes, due to the limited information they contain. The project management tools generate this type of texts and it is necessary to give them the most appropriate treatment to obtain valid and reliable results. In the Xedro-GESPRO tool, based on the analysis of the evidence of human resources’ performance, an evaluation model of labor competencies in projects is generated. This model was implemented without a variant that considered semantic or contextual information for the texts comparison, but the previously mentioned evidences, as well as labor competencies, are short texts. To deal with this problem, it was suggested to use a model based on artificial neural networks (ANN). Specifically, the proposed model was able to capture semantic information from a specialized in project management textual corpus, and improved the current analysis of the indicators that are used for the evaluation of labor competencies. The model not only showed that it is possible to deal effectively with the problems of natural language processing associated with short texts without affecting efficiency and significantly improving the effectiveness and validity of the process but to capture relationship between bad written terms or English words.

Keywords : knowledge discovery; project management; information recovery; short texts; text similarity; PLN; Word2Vec; skip-gram.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )