Instructor: Don Byrd, Visiting Associate Professor of Informatics; Senior Scholar, School of Music

Email: donbyrd@indiana.edu

Phone: 856-0129 (often) and 856-2230 (rarely)

Offices: Music Library, Simon Center (often) (ask for me at the Reference or Circulation Desk) and Eigenmann 1124 (rarely)

Office Hours: 2:15 - 3:30 PM Fri in my Music Library office, and by appointment

AI: Don Reiman, dreiman@indiana.edu

AI Office Hours: 11:30 - 12:30 PM Tues in Eigenmann 930, and by appointment

We meet Mon/Weds/Fri from 1:25 to 2:15 PM, in Simon Center M373. To get to M373, you must go into the Music Library; then go up two floors.

Course Description

In these days of iTunes and pocket-sized MP3 players, people listen to music in digital form all the time, but many exciting possibilities for computer handling of music are still in various stages of research and development. Systems exist now that try to identify–from databases of, in some cases, many hundreds of thousands of recordings–music heard over a jukebox in a noisy bar and transmitted via cell phone, or music hummed into a microphone. Other systems can search a database of scores or MIDI files for a pattern of pitch intervals or note durations, for a chord progression, or for music in a given genre. The Variations2 system, which lets you search for recordings by Martha Argerich of Chopin pieces in minor keys, is now in regular use at IU, both on campus and on computers in student and faculty homes. And concerts have been given in which computers accompany live musicians by playing a predetermined score but “following” the musicians' lead, or by “improvising” freely or with tiny snippets of music from a database of hit songs.

There is a vast difference between how an audio recording on one extreme, and a score on the other, must be represented in a computer. An audio file is organized in more-or-less the simplest possible way, namely as a sequence of (typically) millions of numbers, each representing the loudness of the sound for a tiny fraction of a second. A score file–while generally much smaller–is enormously more complex, and understanding it involves concepts like voices, staves, and transposition, as well as a long list of symbols. (A MIDI file is somewhere in the middle.) But does this vast difference matter if you simply want to find riffs that sound like a given example, flashy trumpet duets, or the name of a tune on the radio? Yes.

In this course, we will consider, among other things:

• how musical information can be organized and represented
• how musical information can be manipulated and displayed
• why it's hard to interpret written music, and why it needs to be interpreted at all
• why it's hard to transcribe performed music into written form
• why some music sounds like other music when you might not expect it, and vice-versa
• how searching for music via content (music information retrieval) works
• how searching for music via metadata (library-catalog style) works

In order to keep in touch with reality, we will also listen to and look at real music as much as possible, and in as wide a variety of styles and genres as possible, but emphasizing those students are interested in.

Course Goals

Music notation of any kind–for example, guitar tablature, Javanese gamelan notation, or conventional Western notation–is dramatically different from the sounds it corresponds to. One of the main goals of the course is for you to understand the profound implications of that fact for almost any application of computers to musical information. (For MIDI fans, piano-roll notation, as seen especially in sequencer programs, is analogous to MIDI; both are somewhere between notation and audio, and we will discuss all three forms.) Other objectives include studying the representation of music and representations of music; learning what the state of the art is for music-seaching programs; and seeing what can be done to explore musical issues (whether theoretical, musicological, music-psychological, or other) with software and databases that are available or becoming available now.

But these are just the technical goals; the real goals are something deeper! Specifically, I'd like students to gain:

1. The ability to evaluate computer music systems: how impressive or unimpressive they really are; how are they related to other systems/task solutions.
2. An appreciation for the subtlety and complexity of our own perception of music; that is, the ability to evaluate human music systems.

Prerequisites

Readings

There is no textbook. Readings will either be available on the Web or on reserve in the Music Library, or I may hand out copies.

Course Outline

The following outline of major topics is approximate; the order of topics and time allocated to each is subject to change.

I. Introduction

Week 1-2. Course Logistics, Introduction, and Motivation (1 session)

   Organizing and Searching Musical Information: A Whirlwind Tour (3 sessions)

   Advanced Motivation: The Music We Like and What's Distinctive About It (1 session)

II. Organization of Musical Information

Week 3-6. Representations of Music and Audio (10 sessions)

   Audio; Acoustics and Psychoacoustics; Formats; Lossless and Lossy Compression

   MIDI, Synthesizers, and Sequencers

Week 7-8. Music Notation; Encodings of Music Notation; XML

   Music Collections: Available or Not, Free or At Cost

   Software for Handling and Converting Encodings/Representations; AMR and OMR

   Hands-on: AMR and OMR

III. Finding Musical Information

Week 9. Metadata (the old way), Content, and "Collaboration"

   Browsing, Searching, and Filtering; Two Kinds of Searching and Three of Filtering

   Mid-term Presentations (2 sessions)

--- SPRING BREAK ---

Week 10. Browsing vs. Searching/Filtering; Searching from Shazam to OMRAS

   Music-IR Evaluation: Precision, Recall, & Relevance Judgments

IV. Musical Similarity and Finding Music by Content

Week 11.

   OMRAS Polyphonic Audio Search via Harmonic Distributions

   Hands-on: Symbolic Searching via Humdrum, Themefinder, NightingaleSearch

   IR vs. Digital Libraries; IR as finding aid; Barlow & Morgenstern

   Music-IR Evaluation: the Cranfield model; TREC and MIREX

V. Finding Music via Metadata and Doing Stuff with Music the Moment It's Found

Weeks 12 & 13. The Full Range of Music-IR Tasks General and Music-specific Metadata; Bibliographic Searching, Filtering, etc.

• Music Recommender Systems
• Similarity Maps, Visualization, and Finding Music by Browsing
• Variations2, iTunes, and Music Libraries; Library Metadata: V2 and FRBR vs. MARC
• Genre Identification and Genre Classifications; iTunes, AllMusic Guide, Ishkur, MIREX, etc.
• Real vs. Fake Reality: Expectation and Perception with Sponges, Dinosaurs, and Music
• Real-time Compositional Applications (e.g., accompaniment)
• Determining Tempo and Beat; Digital Performer and research systems

Week 14-15. Intellectual Property Rights (& sidestepping via XOR?), etc. (1 session)

   Final Presentations (4 sessions)

   Review (1 session)

Course Requirements and Grading

Miscellaneous

If you have a documented disability and anticipate needing accommodations in this course, please make arrangements to meet with me soon.

If you have any comments, questions, or suggestions, please see me during office hours, make an appointment, write me a note (anonymously if you like), or send me email.

University policies on academic dishonesty will be followed. Cite your sources. Students found to be engaging in plagiarism, cheating, or other types of dishonesty will receive an F for the course. For further information, see the IU Code of Student Ethics at http://campuslife.indiana.edu/Code/index1.html .

Late work will not be accepted without prior arrangement for compelling reasons.