Pythonic Metal
summary
Counts on a scraped darklyrics corpus of ~200k songs (~110k English) produce concrete results: a metalness (log ratio) score ranks “Burn” as the top metal word, TF‑IDF + cosine similarity groups bands and reveals subgenres, and clustering of band‑mean vectors recovers meaningful clusters. A 4‑word Markov model generates plausible lyrics quickly, while an RNN trained 40 hours on a laptop improves repetition only marginally. Code and slides are online and the generators work as claimed.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:35Exploratory Data Analysis
- 2:27Overview
- 8:29The Bag of Words Approach
- 9:23Major Themes in Heavy Metal
- 11:48Sentiment Analyzers
- 23:24Bayes Theorem
related talks
15mTalk Less, Graph More: NLP, Networks and Hamilton Musical
58mHow to use ML NET to write crushing metal riffs
31mHow to Use Machine Learning and .NET to Write Crushing Metal Riffs
31mAnalyzing Rap Lyrics Using Python
28mLyrics Generator + Results: AI Usecases for climate change
23mIdentifying Repetitive Songs using LZ Compression
41mMusic and Machine Learning
40mBuilding A Folk Music Recommendation System with LLMs