Research use cases for J-DISC MIR tools

MIR tools – current capability in the field or of J-DISC MIR

1. A highly regarded recording engineer has stated that he did not actually do many jazz recordings credited to him by a canonical jazz label. But he did not specify which ones he did or did not record. Using a given recording studio ambiance quality as a ground truth, we could search all recordings on that label during the pertinent time period and determine which ones he actually did.  Further research could then compare A&R or engineering quality at the other sessions, and ask what the significant differences might be. Still further but feasible work would be to characterize the typical recording ambience at a given level of sound technology—and thus to be able to pinpoint, within a few years, the date of a given recording as a classifier for research purposes.

2. Find every recorded track, from a sample of 20, in which a given instrument takes a featured solo improvisation. If the instrumentalist were known, it would be possible to identify where to find that artist’s personal statement, and avoid listening to tracks where s/he was only accompanying others.

3.  This task follows from #2, but for a larger sample: Find every performance in which an artist takes a featured solo improvisation, and is not simply an accompanist, in the entire recorded output, even in the thousands of songs, where that artist is nominally a part (something traditional discographers—human listeners—would never be able to work through).


MIR tools – projected for next phase of MIR work

4. A library patron working with a large data set of audio files can use J-DISC MIR tools to obtain information on various attributes of their data, such as information on pitch, rhythm, timbre, and performance attributes.

5. Find all bass and drum solo breaks with Paul Chambers as bassist, within a specified range of dates, or with a particular producer.

6. Taking #5 further, analyze all Paul Chambers bass solos from a specified date range, and plot his performances over the compass of the instrument. Determine if there are relevant differences in his performances, and whether they correlate with any other data points. One result might be that he is more likely to play more material in the highest registers when tempo is over 180 bpm.

7.  Find every instance of a carefully defined and modeled expressive trait, such as a glissando or bend, used by modern jazz singers, active in New York in the 1950s and ‘60s, who seek to imitate the technical prowess of instruments, and generate graphics showing the typical musical “shapes.” Develop resulting visualizations to instruct fledgling performers in their use.

8. From a representative sample of jazz recordings, compare the use of an innovative melodic trope, the fourth triad (for example, d-g-c, e-a-d, or all transpositions), before and after 1961, when John Coltrane introduced it and a system of harmonic progression based on it. It is expected that it would be used far more after Coltrane’s recorded oeuvre introduced it, but with a lag of several years as other musicians absorbed and developed the system. The result would be to provide direct evidence of Coltrane’s influence, and to suggest ways to study the dissemination of musical ideas through recordings.

9. Find all jazz recordings made within a 6-month period. Using J-DISC text search, indicate stylistic diversity, patterns in personnel, instrumentation, repertoire, (including repertoire titles themselves). Use J-DISC MIR tools to support investigation of rhythm accompaniment styles or basic harmonic format. As a next step, compare this sample with an appropriate 6-month sample earlier, or later; or compare these trends with respect to jazz recordings in a different geographic location at the same point in history.

10. Access all of Charlie Parker recordings of How High the Moon/ Ornithology, then run queries identifying elements of his approach to the song such as standard material, experiments and recombinations of standard material, or “singularities” (unique and original creative moments.)

11. Assume that it is possible to profile a given artist’s solo style by their timbre, attack, or repeated melodic material. With a single query, a researcher or fan would be able bring back all instances available on the Internet in which two musicians played with each other, based on their sonic profile alone and not dependent on any text annotation of the individual resources.

12. For musical repertoires that do not use a musical score as a central performance document (such as jazz and improvised music, much electronic music, and some non-Western repertoires), MIR tools can provide valuable information about formal structure and data for comparative musical analysis. For instance, a researcher could examine the frequency of motivic patterns across several performers to determine what material might be widely used in common, and what material is unique to a particular performer or situation. Taking the concept in reverse, it would be possible to produce graphic “listening scores” of performances, which might illuminate features of the music not easily caught in the moment of local listening.

13. For musical repertoires that do use musical scores as a central performance document, such as classical repertoire, MIR tools could provide interesting data for analysis on how the score (a standardized set of musical performance instructions) is actually brought to life as a musical performance (since performers almost inevitably add their own expressive features in realizing a notated score, in all forms of music). Work has already been done in comparative mapping of tempi by different artists in performance. MIR tools would open up further possibilities in this type of research.

14. Find, within the full available corpus of a trumpet player’s recordings, the exact repeated elements among the melodic fragments or riffs he or she produces in their solo improvisations, and thus highlight in turn which elements were created spontaneously within the improvisational process.

15. Using the full available corpus of two saxophonists’ recordings, in cases where one was said to be derivative of the other, determine what actual melodic material they both used, identify portions unique to each one, and thereby gather evidence to support evaluations of the artistic originality of the “junior” saxophonist.

16. Numerous private recordings from the loft jazz scene in New York during the early 1970s have been passed down to the artists’ surviving family members or younger participants. No one knows the exact venue and personnel of most of the tapes. Though it is clear there is overlap among the individual recorded sessions, no single party has all of them. After digitization, the solo personnel and combinations could be identified based on ground truth samples of those artists’ commercial recordings. In addition, the duplicates could be identified and differing venues coded by the ambient sound, ranked and selected according to quality, and the exact location and date could be correlated with personnel pending further historical research. When this process was complete, the heirs would be prepared to issue the material commercially in a single package or branded series of issues.

17. Using discographic metadata to identify every recording of a single, widely recorded big band arrangement, or item of standard New Orleans Ragtime repertoire, each version could be searched and marginal variations in the performances identified.  With the ability to pinpoint these variations, researchers could ask whether the arrangers or bandleaders were more concerned with adding new solo improvisations or revising the written portions, or explore many other details of performance such as tempo, inflection, and intonation.

18. Using identification and analysis of song forms by their bar lengths, research within a large set of given recordings can uncover patterns in stylistic innovation and ask how and when innovations, carefully defined, might be adopted. For example, it is assumed that early modern jazz composers of the 1940s adhered to the standard 32-bar song forms divided into 8-bar units favored by Tin Pan Alley and Swing composers as a basis for improvisations. Progressive jazz that emerged slightly afterward was marked by bar lengths and song formats that included odd numbers of bars or unconventional combinations of bar lengths.  After choosing a large corpus of jazz recordings made within the relevant time span, researchers could explore how widespread the change to uneven bar lengths was after the first recorded innovations were disseminated, and how quickly or slowly the change was adopted throughout the whole profession.

19. Currently, many music scores are slowly being digitized as score images. The next phase should include the actual encoding of the data from the scores (using such tools as the Music Encoding Initiative, MEI). This could provide a researcher with a large data set of both music notation (via the encoded score data) and audio performances (via the MIR tools analysis of the audio files). A researcher could cross-reference melodic motives in Mozart piano sonatas, for example, with historical patterns in the live performance of those motives. It might then be explored whether there are certain melodic figures uniformly treated with rubato.


This entry was posted in J-DISC Use cases. Bookmark the permalink.