As a performing art, music is deeply rooted in events occurring over time. Examples of music events are concerts, rehearsals, and recordings. Therefore, when modeling and analyzing music information, an event-driven approach can be necessary. Our goal is to discuss some preliminary steps towards the development of an ontology for modeling and reasoning over music events. The motivations for this study are grounded on musicology with the objective of supporting musicologists by providing computational models and tools to create and analyze datasets. Because of the large scope of this investigation, we focus here on performances. In particular, we look at the connection between performances and their reception, first, because musicologists are often interested in studying how repertories are critically received, second, because the intersection between them ofers interesting research challenges.
CEUR ISSN1613-0073
Our knowledge about music, as a performing art, is deeply bound up with individual moments.
Concerts are events. So are rehearsals and recording sessions. However, musical events are
found in other contexts too: the sum of moments that constitute the act of writing music, its
publication and arrangement, travels of the musicians/composers from one place to another, and
the critical analysis of music by journalists and scholars can all be considered events. Viewed
from the standpoint of knowledge representation and data modeling, these are both a challenge
and an opportunity, particularly once we see these in relation to the discipline of musicology,
which is turning increasingly to digital archives for information about music, in all its riches
(see, for example, [
The purpose of this paper is to present preliminary steps toward developing an ontology that
can model performances and their reception, making music data on (critical and musical)
interpretation computationally explorable. While in fact there are existing research and application
results for the representation of music (see [
According to the Britannica Encyclopedia,1 a performance is a “[...] process during which
musical ideas are realized and transmitted to a listener.” The crowd who attended the concert by
the great virtuoso pianist Lang Lang held at Carnegie Hall in New York City on March 12, 20252
of course went to a generous selection of pieces by composers like Chopin (among others). But
they also went to hear Lang Lang’s particular interpretation of these pieces, just as on another
night they might have attended a performance of Shakespeare’s Hamlet with the expectation
that would witness something exciting, or new, or unique reading of the author’s (or in this
case, composer’s) written text. In other genres and settings audiences might have diferent
expectations for a given event. In the case of a performance by a jazz trio, the “musical works”
performed would matter almost not at all (few such concerts would announce the program,
although the performers and audiences might have fun guessing how a given tune or chord were
transformed along the say). Similarly, diferent experts in Hindustani music might announce a
1https://www.britannica.com/art/musical-performance.
2“Lang Lang, Piano”: https://www.carnegiehall.org/Calendar/2025/03/12/Lang-Lang-Piano-0800PM.
raga for a given evening concert, but most members of the audience would be listening not for
a “work”, but instead a rendition according to the rules of a given genre or master’s style. In the
case of Early Music (roughly: European music up to about the year 1600), which is at the heart
of our interests in musicological research, performances are perhaps best called “realizations” of
texts that leave many decisions to the musicians themselves–choices of instruments and voices,
application of ornaments, any many other expressive details that were never part of the original
notation. Clearly this music changes from event to event, in ways that a scholar like Taruskin
was keen to point out [
This is all to say that musical events are complex objects that require care if we are to model them in ways that make them accessible to machines no less than humans. From a general musicological perspective, studying the changing presentation of musical works helps us detect new meanings latent within them, and understand how they relate to a particular set of circumstances in their reception, which might vary according to historical moment, place, or audience as well as the nature of the sources telling us about it. An ontology that helps point to particular performance of works will be a key part of connecting information on their creation with the qualitative dimensions of how they are received and understood.
Semantic Web (SW) languages and technologies seem a good way to advance these aims.
At a formal level, reasoning systems can help us detect inconsistencies in the data. But they
can do much more than this, for they also allow us to infer things that are only implicit in the
data, further expanding their analytical potential. To make a simple example, the performers
of the rendition of a work should be present at the same temporal span of the rendition itself.
SW ontologies can also support the alignment of data to the FAIR principles [
What would the requirements of such an ontology look like for the performing arts in
general and musical performances in particular? An ontology for modeling and connecting
performances and their reception should enable the representation, querying, and reasoning
of (at least) the information presented in the Table 1 in the form of (functional) requirements.
Specifically, R1-R5 are basic requirements for performances, performers, and musical works.
3See [
R6-R9 refer to reception. Note that R9 may require reasoning about contradictory perspectives on the same performance. This is common in reception analysis, where critics may strongly disagree with each other. Therefore, an ontology capable of satisfying this requirement must be able to reason over these types of claims without encountering logical inconsistencies.
There exists diferent ontologies for music data management and knowledge representation
(for a review see [
Polifonia Ontology Network. A number of SW ontologies for the representation of music
have been presented in the context of the European (H2020) project Polifonia [
For the modeling of music events, MM covers (at least) the classes: (i) CreativeAction for “[...] activities related to the creation, production, or development of artistic works, including music” (OWL file, see footnote 6); (ii) MusicalPerformance for “[a] live or recorded rendition of a musical composition. It encompasses the act of performing music, whether by musicians, singers, or other performers, to convey the intended [...] interpretation of the composition.”
The formal axiomatization of the MM ontology provides only a few axioms for characterizing the intended meaning of these classes. For example, creative actions involve agents who are music artists, and musical performances are intended to “create” musical entities. At the data level, various relations can be used, for example, to model the place or time related to a specific performance, among other information.
The Doremus project. The ontology of DOREMUS7 focuses on the representation of data
relative to classical music [
The representation of performances is done through the general LRMoo’s class of F31
Performance. Following the documentation of LRMoo, this class “[...] comprises activities
where an instance of [F1 Work] is presented or communicated directly or indirectly to an
audience, such as a theatrical play or musical work” [11, p.7]. Similarly to LRMoo, performances
in the DOREMUS ontology can be represented at diferent levels of granularity; e.g., a larger
concert can include various performances of diferent pieces. The axiomatization of the ontology
is mainly limited to taxonomic relations, whereas at the data level, by relying on both LRMoo
and CIDOC, it provides various relations for the specification of data instances.
Music Ontology. The Music Ontology (MO)8 is likely one of the earliest projects for the
application of the SW in music [
The modeling of events in the MO, among which performances, is based on the Event Ontology.9 Similarly to the ontologies previously presented, MO provides only a simple taxonomic representation for most of its modeling elements with various relations used at the data level; these allow the representation of the subevents in a larger event, the time (or place) when (where) an event occurs, the agents involved in the event, etc. 7https://data.doremus.org/; it includes a SPARQL endpoint to access data. 8http://musicontology.com/; ontology files: https://github.com/motools/musicontology. 9http://purl.org/NET/c4dm/event.owl.
Remarks. The three surveyed ontologies ofer three alternative but (apparently) similar
perspectives on the domain of music. In particular, the core distinctions of LRMoo are adopted
in both the DOREMUS ontology and the MO. In contrast, MM is grounded on its own conceptual
foundations, partly motivated by ambiguities in LRMoo [
In relation to the requirements in Section 2, in the cases of R1-R5, the ontologies appear adaptable of satisfying them. On the other hand, the connection between works, performers or performances and the dimension of reception, as required by R6-R9, remains scarcely explored. This is even more evident for requirement R9, as existing ontologies lack the formal means to automatically reason to compute claims that agree or conflict with each other. 10
In the Polifonia library, some work in this latter direction has been presented in [
As noted, the MM, MO, and the DOREMUS ontologies provide only light-weight
formalizations, mainly limited to taxonomic representations, domain/range constraints for relations, and
a few other types of axioms. This design choice may be motivated by pragmatic
considerations: simpler ontologies often scale better in data-intensive applications and are more tractable
computationally [
We report in this section on two datasets to analyze the types of information they represent, their level of granularity, and the underlying models they adopt (when available). Our goal is to understand the kinds of information these datasets prioritize for preservation, the requirements 10These considerations hold also for other ontologies in the state of the art, such as the Performed Music Ontology (https://performedmusicontology.org/), which are not presented in the paper due to space limitations. they reflect, and their respective similarities and diferences. As we will see, the datasets constitute an application scenario for an ontology of music performances and their reception, enabling if not their full integration, at least a uniform access layer across them.
We focus on datasets developed by academic or cultural institutions for purposes of scholarship and dissemination, namely: (i) the Early Music Concerts Database11, and (ii) the Carnegie Hall knowledge-base.12 While many other datasets on musical performances exist – examples are JazzCats13 and Dezède14 – our selection is motivated by four main factors. First, one of the datasets (the Early Music Concerts Database) has a focus on Early Music, which aligns with our research interests. Second, the datasets, especially the Carnegie Hall knowledge-base, can be related to (textual) information about the reception of performances. Third, these resources remain relatively underexplored in the broader research community; by highlighting them, we aim to draw attention to their potential for scholarly investigation. Fourth, the authors of this paper are engaged in ongoing collaborations with the producers of the datasets, who have expressed interest in aligning their work with other initiatives.
We will make examples relative to the composer Josquin des Prez (second half of 15th beginning of 16th century), who was no less a pivotal figure in his own time than Mozart or Beethoven were in theirs. And among his works, one piece in particular (a work for four-voice choir known by its Latin incipit Ave Maria …virgo serena) can serve as a good point of inquiry for our eforts. The piece brings with it all of the usual complexities of works from the years around 1500: none of the existing sources for it can be conclusively tied to the composer himself, nor does any of them spell out much more than the notes, rhythms and general alignment of words and music. Many things are left to the discretion of individual and their directors, including accent, phrasing, pacing, and other expressive details. We know that these sorts of things must vary from one performance (or rendition) to the next. But to make sense of these (and to connect them with critical assertions about those events) we need to be attentive to the data. What do these look like? Here we consider two digital representations of two concerts that featured Josquin’s piece: the first concerns a concert held in Brussels in 1946; the second an event held in New York City in 2015.
Concerts database. The Early Music Concerts Database “[...] catalogs concert programs of Medieval and Renaissance music from roughly 1915 through 1960. It aims to answer questions of how the early music canon was formed and how scholars, performers, and audiences alike rediscovered long-lost repertoires” (from the Website, see footnote 11).
Tables 2 and 3 (partially) show examples of records in the relational database that have been extracted from performances’ programs. At the current state, the documentation exposes two tables: the “concert” table (e.g., Table 2), providing information on performances, and the “work” table (Table 3) giving information on the pieces played during performances. In the examples, the tables show information related to the performance of Josquin’s Ave Maria … virgo serena. 11https://concertsdatabase.org/ . 12https://data.carnegiehall.org/sparql/. 13https://jazzcats.cdhr.anu.edu.au/ 14https://dezede.org/. Dezède will be considered in more detailed applications of our study. Carnegie Hall knowledge-base. The RDF knowledge-base developed and maintained by the Carnegie Hall Data Lab stores data about performances that took place at the Carnegie Hall. The data is organized according to various SW resources assembled together in a light-weight ontology (see remarks below).
Tables 4-5 (partially) show an example of how the data are organized. In particular, the first table represents a concert (instance of schema:Event from Schema.org15) that took place in (schema:location) the Weill Recital Hall of the Carnegie Hall on (schema:startDate) May 6, 2015.16 During the concert, various musical pieces were played, each performance being represented as a sub-event (schema:subEvent) of the whole concert. One of these was the performance of Josquin’s Ave Maria (identified by the Internationalized Resource Identifier (IRI): http://data.carnegiehall.org/works/93901), see Table 5. As it can be seen from both tables, for each (sub-)event, the knowledge-base specifies various information, including the rather surprising fact that this particular group (a chamber ensemble from Florida State University) was not a quartet of singers (as Josquin might have imagined) but instead a quartet of saxophonists. Of course in this rendition the piece would be stripped of its sacred text, but might gain much by way of expressive possibility and timbre. It is hardly historicist in its approach, but it is nevertheless rich in what it tells us about the place of Early Music in modern music academies and in concert life.
By further exploring the data, one ultimately finds that the knowledge-base refers to two diferent works by Josquin with the same incipit Ave Maria, each with its own IRI: work/3527517 and work/9390118 (showed in Table 5). The latter is an arrangement by Dave Wozniak, a noted saxophone soloist and conductor. Although this suggests that work/93901 is an arrangement of 15https://schema.org/. 16The data is available here: https://data.carnegiehall.org/events/54454/about. 17https://data.carnegiehall.org/works/35275. 18https://data.carnegiehall.org/works/93901. Josquin’s Ave Maria, this choice is not explicit in the formal representation of the data. Nor is this the end of the story, for there are in fact other compositions by Josquin that begin with these words.19 These sorts of problems are not infrequent when working with Early Music. Nevertheless, the Carnegie Hall knowledge-base (unlike the Early Music Concerts Database) does not provide enough information to unambiguously identify the works heard there. Remarks. The datasets ofer distinct perspectives on performances. From a knowledge representation standpoint, they difer in (at least) modeling formalism and granularity of detail.
A relevant feature of the Carnegie Hall knowledge-base is its fine-grained modeling of event structure, including the representation of sub-events. This enables the explicit representation of information applicable either to the entire event or to specific sub-events. However, although the knowledge-base makes use of SW languages and technologies, it lacks a formally specified ontology. While the data does rely on existing and custom ontologies, no comprehensive ontology is published to characterize the core domain concepts and their relations and formal 19It was probably the famous Ave Maria …virgo serena but it might also have been Ave Maria …benedicta tu. axiomatization. As a result, domain assumptions – such as that performers’ roles must be iflled by human agents – are not formally encoded. This absence of a logical theory behind the knowledge-base limits support for automated reasoning such as consistency checking or inferring implicit facts. For example, inferences that make information relative to a whole event propagating to its sub-events – e.g., if the whole concert takes place in a location at a certain time, all the performances within the concert must have the same location and must occur within the time-frame of the concert – cannot be automatized.
In the Early Music Concerts Database the granularity of information difers with respect to the Carnegie Hall knowledge-base. Performances are modeled as “atomic” events – each representing a single concert, with associated data and a list of performed pieces. Also, it emphasizes scholarly details such as the modern edition or source of the performed pieces with less focus on performer-level granularity.
An ontology that can account for musical events along with information relative to the performed music pieces, performers, venues, source editions, etc. can serve as a common layer enabling federated querying or even the integration of the datasets. Also, for musicological research, accessing these datasets in a uniform way can open new possibilities for comparative analyses. Scholars could explore, for example, the evolving performance history of specific pieces across institutions and decades; trace the careers of performers or ensembles through diferent archives; examine the re-emergence of Early Music through the analysis of performances programming; or analyze the difusion of editorial traditions for Renaissance and Baroque works. In a more general perspective, the ability to query multiple performance datasets uniformly can support data-driven investigations into canon formation, stylistic evolution, and historiographical trends in music performance, just to mention a few cases.
We have presented through the paper some preliminary steps towards the development of an ontology of music events focusing on the case of performances and their critical reception. The motivations for this study are related to music criticism: if data about music and reception is formally represented, the data can be made computationally processable to support both qualitative and quantitative musicological research.
As seen in the previous sections, a number of ontologies for music exist. However, these
exhibit limitations in terms of either their formal representation or their conceptual expressivity.
In particular, they lack the mechanisms needed to relate musical works and performances to their
reception, such as critical discourse, reviews, or even audience response (as in the infamous case
of the near riot that erupted during the 1913 premiere of Stravinsky’s Rite of Spring) that may
influence the way we think about particular pieces. In our view, this gap underscores the need
for further research into ontological frameworks capable of connecting music and reception.
The modeling of reception is particularly challenging, for example, because one has to be able to
represent alternative and possibly contradictory information on the same entity, but also because
one needs to develop shared vocabularies to represent the claims expressed in reception texts.
This is complex because of the conceptual plurality and subjectivity of criticism. Future research
will attempt to refine existing ontologies like OMAC [
Various challenges at the intersection between musicology and knowledge representation remain open. To mention a case, the examples discussed in the paper come from the Western Art Music tradition. Here the work concept is key, especially to the “core” concert repertory (say, from Bach to Brahms). These kinds of concerts are events involving the interpretation of some abstract thing we call in fact a work, which in turn is preserved as a kind of text (score). Of course it is more complex than this, since scores (and therefore the works they preserve) are surprisingly unstable. Composers are not always aware of everything that a performer needs to do; editors adapt and interpret existing scores for new readers.
Once we venture beyond the core repertory, the notion of work, and thus what we mean by performance, becomes cloudy. To mention just some examples, some avant garde and experimental pieces do not have fixed scores (and so performers are sometimes composers). Others do not have a text other than the sound recording or digital file that preserves them (and so an audition of these does not involve a “performer” in the conventional sense at all). For Early Music things are similarly hard: since written sources often left a lot of things up to performers; sometimes we cannot tell what is an original piece and what is its derivative. In addition, one needs to be ready to deal with the many oral traditions that do not have any sense of a written work at all, from ragas to blues songs.
Existing ontologies often treat concepts such as work and performance as self-evident, relying on implicit understandings within particular user communities or the functional needs of specific applications. From a scholarly standpoint, however, this is insuficient. Hence, an ontology that aims to serve the needs of musicological inquiry must engage seriously with the conceptual and historical intricacies of music and performing arts in a more general sense. Developing such an ontology requires not only formal precision but also close collaboration with music scholars since only through such interdisciplinary engagement we can produce models that are both computationally viable and epistemologically sound.
The core ideas of this paper were developed during Sanfilippo’s and Freedman’s research stay at the CESR in Tours in the spring of 2025. The authors would like to thank their colleagues for the discussions that led to this preliminary work.
During the preparation of this work, the authors partially used DeepL Write in order to: Grammar and spelling check. After using this service, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content.