Difference between revisions of "Music 254"

From CCARH Wiki
Jump to: navigation, search
(Introduction)
(Introduction)
Line 18: Line 18:
  
 
=Introduction=
 
=Introduction=
Music 254, offered in the spring quarter, has evolved over more than two decades from a seminar in music representation systems to a project-oriented sandbox for applications involving the use of symbolic (discrete) musical data. Music 253, offered in the winter quarter, delves into the details of the professional music-typesetting program SCORE; the internal format of MIDI files; the Humdrum Kern format for music analysis and visualization; inherent differences in musical-information qualities between sound and notation formats; and data-interchange methods including MusicXML and MEI.  Projects and applications can be developed on the basis of any of the above and often involve information from an additional domain (gesture, audio, et al.). This course has been an incubator for software applications in such areas as polyphonic query, the visualization of tonal harmony, real-time analysis of pre-tonal and tonal music, and music information retrieval (MIR) in Western and non-Western repertories.
 
  
Labs are focused on selected features of the <i>Humdrum Toolkit</i> [http://csml.som.ohio-state.edu/Humdrum/talks/Humdrum.Toolkit/index.html] to design their own research in areas appropriate to their background and interests. Previous programming experience is not necessary, but a working knowledge of musical notation and elementary concepts of music theory is.  
+
==Overview==
 +
Music 254/CS 275B is a seminar offered to graduate students in music and in computer science in the spring quarter of each year. Its overall aim is to enable projects in computational music analysis using symbolic data.  
  
The potential list of topics that <i>may</i> be covered in Music 254 is substantial, but in any given year those most salient to the skills and interests of the class are emphasized. The first month requires regular class meetings, an introduction and preliminary exercises in the use of Humdrum, and individually assigned readings in an area of mutual interest. Individual weekly meetings replace lab time in the last half of the quarter.  
+
The first month requires regular class meetings, an introduction and preliminary exercises in the use of Humdrum, and individually assigned readings in an area of mutual interest. Individual weekly meetings replace lab time in the last half of the quarter.  
  
==Definitions==
+
==Project topics==
Ultimately, all the application areas covered in Music 254 initially use the two large data repositories under continuous development at [http://www.ccarh.org CCARH]. In symbolic music applications, MIR is treated as a specific branch of the digital analaysis of musical repertories.  
+
Projects and applications can be developed on the basis of any well documented encoding scheme and may draw on machine-readable information from an additional domain (gesture, audio, et al.). Music 254/CS 275B has been an incubator for software applications in such areas as melodic search in Western and non-Western repertories, polyphonic query, harmonic visualization, real-time analysis of pre-tonal and tonal music, generation of compositions based on specified styles and/or genres, and AI techniques in author/composer profiling, structural dissection of polyphonic scores, and feature-weighting approaches to MIR (music information retrieval).  
  
The component covering style evaluation and simulation seeks to define those musical features that constitute a particular style of composition and differentiate it from all others. Feature sets, once defined, can serve as a springboard to the generation of new music in the style of an existing repertory.
+
==Prerequesites==
 +
Successful completion of Music 253/CS 275A, Symbolic Musical Information and graduate status or evidence of equivalent preparation.
 +
 
 +
Previous programming experience is not necessary but is an asset.  A  working knowledge of musical notation and elementary concepts of music theory is essential.
 +
 
 +
Familiarity with other recent work in the chosen project area is essential.
 +
 
 +
==Historical background==
 +
Music 254/CS 275B has evolved over more than two decades from a seminar in music representation systems to a project-oriented sandbox for applications involving the use of symbolic (discrete) musical data.
 +
 
 +
Music 253/CS 275A, offered in the winter quarter, delves into the details of representation schemes and code content for programs in music typesetting (SCORE, Finale, MuseScore), MIDI, interchange codes (MusicXML, MEI), and related methods of data acquisition including optical music recognition. A recent syllabus and description are [http://wiki.ccarh.org/wiki/Music_253 here].
 +
 
 +
==Labs==
 +
Labs are focused on learning selected features of the <i>Humdrum Toolkit</i> [http://csml.som.ohio-state.edu/Humdrum/talks/Humdrum.Toolkit/index.html] to design their own research in areas appropriate to their background and interests. As the course progresses, users can focus on the tools relevant to their projects.  They are also welcome to write new ones. 
 +
 
 +
==Data Resources==
 +
Data repositories of verified symbolic encodings include three that have been created locally:
 +
 
 +
* <b><i>MuseData</i></b>
 +
Scores of Western European music--mainly for orchestra, ensemble, quartet, or mixed group--from 1700 to 1850 (e.g. Bach, Corelli, Vivaldi, Haydn, Beethoven, et al.) fully encoded by humans (with lyrics when pertinent) in the MuseData format, with conversions to several other formats.
 +
 
 +
* <b><i>KernScores</i></b>
 +
Scores of Western European music--mainly for piano--(classical period through early twentieth-century, with broad edges for earlier and later music. Optically recognized. Humdrum **kern format native, with on-the-fly conversions to numerous other formats for notation and analysis.
 +
 
 +
* <b>The Josquin Research Project</b> contains scores--mainly choral repertories composed by Josquin and his contemporaries. No lyrics. Encoded by humans for use with **kern and MuseData. Conversions to other formats include MIDI, MEI, et al.
  
 
==The <i>MuseData</i>, <i>KernScores</i>, and Josquin Databases==
 
==The <i>MuseData</i>, <i>KernScores</i>, and Josquin Databases==

Revision as of 23:15, 2 April 2018

See also the Music 254/CS 275b Syllabus.

1 Introduction

1.1 Overview

Music 254/CS 275B is a seminar offered to graduate students in music and in computer science in the spring quarter of each year. Its overall aim is to enable projects in computational music analysis using symbolic data.

The first month requires regular class meetings, an introduction and preliminary exercises in the use of Humdrum, and individually assigned readings in an area of mutual interest. Individual weekly meetings replace lab time in the last half of the quarter.

1.2 Project topics

Projects and applications can be developed on the basis of any well documented encoding scheme and may draw on machine-readable information from an additional domain (gesture, audio, et al.). Music 254/CS 275B has been an incubator for software applications in such areas as melodic search in Western and non-Western repertories, polyphonic query, harmonic visualization, real-time analysis of pre-tonal and tonal music, generation of compositions based on specified styles and/or genres, and AI techniques in author/composer profiling, structural dissection of polyphonic scores, and feature-weighting approaches to MIR (music information retrieval).

1.3 Prerequesites

Successful completion of Music 253/CS 275A, Symbolic Musical Information and graduate status or evidence of equivalent preparation.

Previous programming experience is not necessary but is an asset. A working knowledge of musical notation and elementary concepts of music theory is essential.

Familiarity with other recent work in the chosen project area is essential.

1.4 Historical background

Music 254/CS 275B has evolved over more than two decades from a seminar in music representation systems to a project-oriented sandbox for applications involving the use of symbolic (discrete) musical data.

Music 253/CS 275A, offered in the winter quarter, delves into the details of representation schemes and code content for programs in music typesetting (SCORE, Finale, MuseScore), MIDI, interchange codes (MusicXML, MEI), and related methods of data acquisition including optical music recognition. A recent syllabus and description are here.

1.5 Labs

Labs are focused on learning selected features of the Humdrum Toolkit [1] to design their own research in areas appropriate to their background and interests. As the course progresses, users can focus on the tools relevant to their projects. They are also welcome to write new ones.

1.6 Data Resources

Data repositories of verified symbolic encodings include three that have been created locally:

  • MuseData

Scores of Western European music--mainly for orchestra, ensemble, quartet, or mixed group--from 1700 to 1850 (e.g. Bach, Corelli, Vivaldi, Haydn, Beethoven, et al.) fully encoded by humans (with lyrics when pertinent) in the MuseData format, with conversions to several other formats.

  • KernScores

Scores of Western European music--mainly for piano--(classical period through early twentieth-century, with broad edges for earlier and later music. Optically recognized. Humdrum **kern format native, with on-the-fly conversions to numerous other formats for notation and analysis.

  • The Josquin Research Project contains scores--mainly choral repertories composed by Josquin and his contemporaries. No lyrics. Encoded by humans for use with **kern and MuseData. Conversions to other formats include MIDI, MEI, et al.

1.7 The MuseData, KernScores, and Josquin Databases

The CCARH Lab (Braun Music Center #128) provides support for a wide range of notation and analysis applications. The Center for Computer Assisted Research in the Humanities (CCARH)[2], which hosts it, originated the MuseData encoding format [3] and has encoded approximately 1,200 complete classical works.

The Humdrum Toolkit, initiated by David Huron and officially hosted at Ohio State University, is complemented by additional tools and an extensive collection of music encoded in the **kern format (KernScores [4]). Craig Stuart Sapp is also the developer of the KernScores website [5].

Both of these resources contain extensive amounts of data. The best way to become acquainted with them is to explore them systematically. KernScores will be the more useful for many projects in Music 254/CS 275B, but significant differences in data profiles should be noted. MuseData originates as part-by-part encodings. The data format is intended to support editing, printing, MIDI, and various kinds of re-purposing (re-use). Kern data is through-encoded (all parts are entered at once). MuseData principally contains large-multi-movement works (symphonies, string quartets, ensemble sonatas, operas, oratorios). KernScores principally contains keyboard music. MuseData emphasizes music composed between 1680 and 1850. KernScores has a broader reach, with small quantities of data from the Renaissance, from the early twentieth century, and from folksong repositories.

Beyond these differences in holdings, KernScores offers several paths through its hoildings. The composer and genre lists on the homepage are self-explanatory. At the top of the page, under the search window, the Guided Tour provides useful information on navigation. Data Collection Highlights shows some of the larger constellations of holdings. The Online Humdrum Editor is useful for becoming familiar with the **kern data format used in the database and also for producing short musical examples for print or online use. For full particulars on formats see the CCARH Humdrum Portal.

KernScores has a large number of pre-defined analytical capabilities at the individual work level. Some generate graphical, others alphanumeric displays. The buttons to the left of title listings employ these codes: S=Score, H=Humdrum code, M=MIDI, K=keyscape. A Z, when encountered at the head of a listing, refers to a zip file containing all of the following listings.

On individual work pages, e.g. Chopin's Prelude in C Major, Op. 67, No. 1 (Quasi improvisazione), the user can find translations of Humdrum data ("Data Formats Translation") to notation via several different pieces of software. Of these ABC+ code is among the most robust, MusicXML the most useful for import to Finale and Sibelius. Some of the other formats offer evaluate and/or weigh certain musical features for expressive performance, metrical weighting, and so forth.

The lower portion of individual work pages contains links to produce "keyscape" visualizations of harmonic structure ("keyscapes"), piano-roll views depicting individual voices in relation to each other and showing overall changes of texture, can detect chord roots, and can append them, event by event, to the preceding **kern file.

The Josquin Research Project, directed by Jesse Rodin, contains more than 1,000 pieces of music from the early Renaissance. It was started with the intention of differentiating real Josquin from false attributions to Josquin. The music it contains in in mensural notation. It is designed for musicological queries, which are facilitated through the use of a search panel on the left of the home screen. Music by the composers Dufay, de la Rue, Obrecht, and Ockeghem are included in the data pool.

1.8 The Humdrum Toolkit

The Unix-based Humdrum Toolkit originated in the mid-1980s as a small set of commands that can be chained into series to produce concise information about changes inside an encoded musical score. Although the rudiments seem simple, the diversity of questions that can be posed and answered is potentially immense. Because Humdrum is now supported on three continents, CCARH recently constructed a Humdrum portal [6] to link tools and documentation at diverse locations.

2 Course Topics

See also Recent Projects below.

2.1 Elements of Musical Style

What constitutes a musical style? As a classroom subject, style is often used as a lens through which to differentiate the work of different composers, genres, historical periods, places, schools of pedagogy, and so forth. It is an umbrella term to which almost any aspect of music can be related. While no one denies that the term is vague in its meaning, two individuals who have defined and articulated specific elements of musical style have also provided taxonomies that are inordinately useful in the development of computer programs for musical analysis.

2.1.1 LaRue's Rubrics of Style Analysis

The noted musicologist Jan LaRue (1918-2003)[7] took a top-down approach to the musical work. His taxonomy of musical features embraces a wide range of musical phenomena, most of which are frequently pertinent to classical music of the past four centuries. In his Guidelines for Style Analysis: A Comprehensive Outline of Basic Principals for the Analysis of Musical Style (New York: W.W. Norton, 1970), LaRue fleshes out his 9-part skeleton into three layers of detail.

3 Humdrum Lab Pages

Go to the Humdrum Lab portal for a list of exercises to get up to speed with using the Humdrum Toolkit and other programs/environments to work with Humdrum data files.

4 Recent and current subjects of student and visitor research

Topic ideas and technological tools vary considerably from year to year. The wide scope of projects emerging from Mus 254/CS 275B is best represented by subjects pursued over the past decade.

Degree candidacy and program indicated by initials [CS = computer science, CCRMA = Center for Computer Research in Music and Acoustics, DMA = Doctor of Musical Arts, EE = Electrical Engineering, MST = Music, Science, and Technology program]

  • A geometric approach to content-based retrieval (CS Master's)
  • A database for chord recognition (CCRMA PhD)
  • A keyboard query system for Themefinder (MBA business)
  • A notation system for gesture (DMA)
  • A steel-drum teaching app for calypso (CS Master's)
  • A tabla transcription system (CCRMA PhD)
  • A transcription system for Turkish music (PhD physics)
  • Algorithmic realization of basso continuo using partimento theory (PhD medicine)
  • Analysis of ryoka (CCRMA MST)
  • Automatic accompaniment (CCRMA MST)
  • Automatic reduction from full to piano-vocal score (DMA)
  • Automatic score alignment, audio-symbolic data alignment (EE Master's)
  • Automatic synchonization of music and movement (CCRMA PhD)
  • Beat alignment (numerous, mainly CCRMA)
  • Contrapuntal analysis using the Hausdorff metric (CS MS)
  • Deep learning for analysis of musical structure (CS MS)
  • Deep metric structure (visiting German PhD)
  • Electronic composition for animations (visiting master's from Keio University)
  • Encoding system of music by Hildegard con Bingen (PhD visitor)
  • Encoding Zarlino's treatises with interactive music and sound (visiting Dutch PhD)
  • French chant in the Norman conquest of Southern Italy (visiting PhD from Italy in music/art history)
  • Generation of new Joplin rags (EE Master's)
  • Geospatial cluster analysis of meter in European folksong (Music PhD)
  • Harmonic Analysis and its Visualization (CCRMA PhD)
  • Harpsichord tuning systems in the Italian Baroque (Co-term Music/Math)
  • Interactive Leitmotif analysis tools with integrated score display and sound output (German CS master's)
  • Implementation of Themefinder search tools in the RISM search engine (visiting Swiss PhD in Music/CS)
  • Koto score software (visiting PhD candidate, Keio University)
  • Mapping of drum gesture (PhD CS)
  • Melodic analysis (PhD mathematics)
  • Metrical weights following Neumann, Lerdahl, and Volk (visiting German PhD)
  • Music synchronization in virtual reality (CCRMA PhD)
  • Peer-to-peer query by humming (CS master's)
  • Performance analysis: Bach fugues (CCRMA PhD)
  • Phonology and music-notation mapping (CCRMA PhD)
  • Pitch-tracking for South Indian Classical Music (EE PhD)
  • Probabilistic style-simulation using trees (SSP undergraduate)
  • Perceptual of timbre (CCRMA PhD)
  • Real-time voice pitch tracking in Humdrum (MST undergrad)
  • Rhythmic pattern sort in RISM OPAC data (1.3 million incipits) (PhD collaborator)
  • Drumpfet acoustics (visiting Taiwanese researcher)
  • Searching lyrics in multimedia databases with Mandarin and English (CS PhD)
  • Style judgment by man and machine (EE PhD)
  • Stylistic traits of Pierre de la Rue's motets (Mus undergraduate)
  • Temporal calibration of gagaku (CCRMA MST)
  • Visualization methods for structural analysis (MA mathematics)
  • XML tools for MuseData and Themefinder (visiting Dutch researcher)

5 Ongoing research projects at CCARH


Previous front page: http://www.ccarh.org/courses/254