Currently, the detection of polarities in written texts is oriented to the lexical, syntactic, and semantic levels of the linguistic levels. This doctoral thesis seeks to contribute to the understanding of information from any context to automatically interpret its polarity, detecting characteristics at the lexical, syntactic, semantic, and phonological levels of texts that allow representing polarity in natural language. This research is based on the science of design through the use of a software artifact, to analyze the contribution to the prediction of polarity detection in microblogging sources of phonetic elements (phoneme, phonestheme) and different emotional elements. The valence (arousal/dominance), by using the correlation of each of the elements in an experimental design. The analysis is validated through the data set of the SemEval-2018 Task 1 V-oc, which provides us with the necessary characteristics to analyze the phonetic and emotional elements of the texts in microblogging sources. The results allow determining the polarity using the union of lexical, semantic, phonetic, and emotional elements with an F1 measure close to 80 %. This research uses the symbolic connections of sound to represent and model sound in meaning with the support of emotional elements such as arousal and dominance for classifying texts in Spanish and English. Likewise, these elements can be used to determine the polarity of texts in microblogging sources.