Every day the number of documents produced and shared on the web is growing, and with this increases the need to find techniques and methodologies that are able to recover the most relevant elements based on some notion of similarity. On the basis of this need, this paper aims to present a method for the search of musical text considered in its symbolic aspect (music notes) based on the idea of transforming musical scores into feature vectors and comparing these through measures that define the similarity between the relative representations. To do this, the Cosine Similarity method was used, a heuristic technique generally used for text comparison in data mining and text analysis. This form of information representation is one of the essential components that allowed the design and implementation of an efficient system in terms of response times and accuracy of results.