Procedural Generation of Musical Metrics Based on Lyrics Analysis

Abstract

More than the semantic and discursive components, the musical lyrics often contain other information, that more than with the act of writing, has to do with the act of pronunciation. Assuming that the musical lyrics are written to later be reproduced verbally, there is a caution for this process to pass us something too, completely different from what is conveyed by the lyrics on paper. The synchrony of phonetic and lexical accents of the lyrics with the musical components in which it belongs is a great example of that. In this project, the proposal is to create a system able to return music information for a given lyrics. More specifically, information on the metrics. To this end, I will use the CMUdict, a phonetic information dictionary for English language that contains, for each word, the division of its phonemes with the respective markers related to their stress. The entire operation of the system will be based on Python programming language, having all the code been developed by me especially for the project. For each letter entered, it will run an analysis by verses and each verse will become a metric template. All the verses from the lyrics will be adjusted to each of the templates and will be classified in order to select what is the template that best fits the letter in general. The template with the highest score will be chosen as the metric final structure.