Introduction

MUMT621

Topics

• Course description
  • Overview of selected topics in Music Information Retrieval (MIR) via presentations
  • How to efficiently conduct research and practice presentations
  • Final project, either:
    • Software project with description (2–3 pages) (Repository on Github)
    • Research paper (5–10 pages) (ISMIR style). Submit it to ISMIR?
• Topics: Overview
• MIR resources
• Google Advanced Search
• Wayback Machine
• Literature search
• Filename scheme: [family name of the first author in lower-case]+[2-digit year of publication]+[first real word of the title in lower case], e.g.: fujinaga26optical.pdf
• Backups: Scheduled backup, remote (P2P) syncing, bootable disk
  • MacOS: Time Machine, Resilio, Carbon Copy Cloner/SuperDuper!
  • Windows:File History/Duplicati, Resilio/Syncthing, Rescuezilla
  • Linux: deja-dup, Resilio/Syncthing, Rescuezilla
• Using LLMs Effectively: Prompting, Context, Grounding, and Verification
  • LLMs (Try and compare: ChatGPT, Gemini, Claude, Mistral, Perplexity, Microsoft Copilot)
• McGill Guideline
• Prompt Engineering Guide
• Prompt engineering best practices for ChatGPT (OpenAI)
• Prompt design strategies (Google)
• Prompting best practices (Anthropic)
• Gemini Notebook (formerly Google NotebookLM): source-grounded AI for research and literature review
  • UCLA: NotebookLM for Research (Ingtroduction)
  • NotebookLM In 30 Minutes
• 3Blue1Brown Video, “Large Language Models explained briefly” (2024)

• History (SIGIR 1999: Call, Proceeding; ISMIR 2000)
• Early examples
  • Melodic arch (1996)
  • Melody type map (1999)