Quantitative and qualitative analysis of a multi-genre heavy metal corpus: an NLP-based approach