Prelude for an ontology of documentary musical data. A conceptual analysis of FRBR(oo)

Tracking #: 2384-3598

Authors: 
Emilio M. Sanfilippo
David Fiala
Richard Freedman
Hyacinthe Belliot
Vincent Besson
Marion Lamé

Responsible editor: 
Special Issue Cultural Heritage 2019

Submission type: 
Ontology Description
Abstract: 
Ontologies for the representation and management of both knowledge and data about music are widely used for disparate purposes. In this context, the Functional Requirements for Bibliographic Records (FRBR) and its object-oriented version (FRBRoo) are used to distinguish a musical composition (Work) from its scores (Expression), among other things. Unfortunately, both FRBR and FRBRoo rely on unclear modeling assumptions leading to ambiguous and opaque data models. The purpose of the paper is to dig into some foundational (conceptual) aspects related to the representation of bibliographical musical data. In particular, we first propose a critical evaluation of FRBR/FRBRoo. Second, on the basis of such an evaluation, we depart from these models and propose an alternative approach which keeps some of the useful distinctions in FRBR/FRBRoo while getting rid of their ambiguities. We formally represent our ontology in the Web Ontology Language (OWL) to (ultimately) take advantage of both reasoning and data management methods and technologies.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Reject

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Fabio Vitali submitted on 17/Dec/2019
Suggestion:
Major Revision
Review Comment:

The paper centers around three main themes:

T1) perceived ambiguities and shortcomings of the FRBR (and FRBRoo) conceptual model(s) to describe intellectual works, especially when applied to music and musical scores;
T2) A way to fix such ambiguities and shortcomings by changing definitions and relationships of the accepted model so as obtain a fairly different conceptualization of intellectual works, with only a passing resemblance to the original FRBR (and FRBRoo) original models;
T3) A way to apply such new model and create datasets of statements in the field of music (and, in particular, of printed scores of XV and XVI Century music).

Overall, there is matter to appreciate and reflect upon in this paper. Nonetheless, the good that one could find in the content is severely tainted by a number of issues that prevent me from fully appreciating the work as it was proposed, for a good number of reasons:

R1) I believe there is a fundamental misunderstanding on the basic concepts of FRBR (and FRBRoo) that to me cancels any merit in T1;
R2) I am ready to accept that the application for which they planned to use FRBR does not require some of the features that FRBR supports, but I am disappointed that instead of simply ignoring them, the authors preferred to criticize the FRBR model as ambiguous, confusing and plainly wrong;
R3) This misunderstanding brings some severe problems in their modified model T2. While the modified model has merits, the overall construction lacks on a number of points anyway.
R4) I believe that most of the scholars accustomed to FRBR (and FRBRoo) would use the relevant concepts differently than the authors, in particular Expressions. I also believe the scholars would be right and the authors wrong in this. But mainly I believe that this disagreement would bring a lot of discussions and debates on how to compare and integrate the dataset T3 with other related initiatives.

A fundamental misunderstanding of FRBR was the first thing that I noticed in the paper, where FRBR is systematically described as composed of three levels: "FRBR(oo) make a useful distinction between (at least) three relevant entities for cataloging, i.e., Work, Expression, and Item", p. 2, "DOREMUS ontology has been created by extending both CIDOCCRM and FRBR(oo). It is not therefore surprising that the ontology heavily relies on the triple Work-Expression-Item", p. 5, "We present in the next section a way to deal with FRBR(oo)’s issues while keeping the cut-off distinction between works, expressions, and items." , also p. 5, and "Our approach departs from FRBR(oo) while sharing at the same time some (high-level) commonalities, above all, the triple Work-Expression-Item", p. 12.

The best known descriptions of FRBR use FOUR, and not three, levels. They are named Work, Expression, Manifestation and Item. There is also a widespread acronym for them, WEMI, that is sometimes used instead of FRBR to refer to that conceptualization. In this paper, Manifestations are never mentioned neither in the FRBR model as described, nor obviously in the proposed modification. In fact the word "manifestation" does not appear anywhere in the paper. Yet it is a fundamental level in the structure of FRBR, and in fact one of my objections R3 lies in the improper use of Expressions in lieu of Manifestations in the proposed model.

According to FRBRoo, Manifestations are either "F3 Manifestation Product Types" or "F4 Manifestation Singletons". Given that we are dealing with printed reproductions of musical scores in anthological works, we can reasonably assume that F3 applies. They are defined as "all of the features or traits that instances of F5 Item normally display in order that they may be recognised as copies of a particular publication." Examples are "the publication product containing the text entitled ‘Harmonie universelle’ (authored by the person named ‘Marin Mersenne’), issued in 1636 in Paris by the publisher named ‘Sébastien Cramoisy’ and the publication product containing a modern reprint of Marin Mersenne’s ‘Harmonie universelle’, issued in 1986 in Paris by the publisher named ‘Les éditions du CNRS’, and identified by ISBN ‘2-222-00835-2’"

These words echo almost to the individual property the examples given in the paper for Expressions, such as "Josquin des Prez’s Missa Pange lingua [...] published for the first time in 1539 in the Missæ tredecim quatuor vocum (Nuremberg: arte Hieronymi Graphei, 1539)" and in "the Werken van Josquin des Prez, edited by A. Smijers, H. Vinders, B. Appenzeller, and N. Gombert (Amsterdam: Vereniging voor Nederlands Muziekgeschiedenis, 1922-1965)"

What the authors consider Expressions are, according to my humble understanding of FRBR (and FRBRoo), exactly Manifestations. And therefore we are left without what in FRBRoo are Expressions, making me wonder whether the authors either a) do not have entities to be usefully identified as Expressions, or b) call Works those entities that elsewhere would be called Expressions.

In the words of FRBRoo, "Expression encompasses [...] the specific words, sentences, paragraphs, etc. that result from the realization of a work in the form of a text, or the particular sounds, phrasing, etc. resulting from the realization of a musical work. The boundaries of the entity expression are defined, however, so as to exclude incidental aspects of physical form, such as typeface and page layout for a text". In the musical field, they obviously correspond to scores, recordings, and videos of performances, etc. Additionally, each different content of different variants of the same Work generates a different Expression. For instance, drafts, edited publications, transpositions, transcriptions, etc. that could still be traced back to a single idea of work, would correspond to different Expressions of the same.

I understand that the authors do not deal with registrations or videos of performances of the musical scores they handle, so these types of Expressions are not an issue. But what about Expressions such as variants of the same score, e.g., early drafts, modified publications, transpositions or transcriptions for different instruments? That the authors do not have this kind of situations, is hard to believe. That they choose to assign them to separate Works, is hard to accept. I suspect the authors use a disagreeable definition of Work that includes what I would consider Expressions of the same Work.

This brings the discussion to the definition of Works. The authors find imprecise and ambiguous the definition of Works as "a distinct intellectual or artistic creation", because "It should be clear that two different interpretations of Work are at play: the first one, according to which a work is a cataloging entity; the second one, for which a work is an idea in someone’s mind. These two interpretations cannot be confused and need to be carefully distinguished if one aims at modeling data in a transparent way" (p. 3). The solution proposed by the authors is to use only a specific restriction of the original meaning for the term Work, meant for existing works only: "The proposal presented in this section departs from FRBR(oo) and builds on library science approaches for the modeling of what we call documentary work. [...] Documentary work is a conventional entity that is intentionally created for cataloging purposes, namely, to organize, store, and retrieve data, among other applications. A documentary work is therefore created a posteriori by abstracting from specific and possibly multiple expressions [...] or items".

I am not scared by this choice. It is not, of course, that I agree with the authors about the shortcomings of FRBR(oo) in defining Works, quite to the contrary. But, provided that the authors are content with defining a restriction of FRBR(oo) Works, and call it differently than Work, I am just fine: it does not mean that the authors demolish FRBR(oo) model, but just that they felt the need to carve out of it a cozy and more comfortable model for their own purposes and their needs, which is not illegal nor improper. Good for them.

Of course, we could complain about the complacency with which we are explained that "This view is unfortunate: first, if works are introduced to classify expressions, it cannot be the case that their existence precedes the existence of expressions or that works exist without expressions; second, it is unclear whether FRBR(oo) distinguish between a work and its originating idea(s)." and that "This view in turn now takes us from a problem of data modeling to one of the genesis of artistic work. The idea that works precede their expressions is a Romantic one indeed, aligning as it does with ideas of genius, inspiration, and the Ideal Work."

Coming from a field where Expression-less Works are frequent, I find the notion that this is all a misunderstanding bound in Romanticism a fairly funny conclusion indeed. In many countries (especially based on Civil Law) it is quite frequent for the Legislative power to generate acts that deal with very general concepts (e.g., "a new family of research grants with a budget of X Millions euro is established for the next fiscal year"), and then leave to the executive power the details of how to make use of such concepts (e.g., when is the deadline, who the submission should be addressed to, what type of documents should be submitted, how is the evaluation handled, etc.). In the legislative act, the document detailing these aspects is explicitly referred to with a pretty normal legal reference to a Regulation of the Executive power that, at the moment of enactment of the law, does not exist yet, but exist in abstract as a future document. This is not an open or vague reference to the hope that such a document will exist sooner or later, but a very normal reference that, once such document is created, is indistinguishable from any other legal reference of the same kind. The Work of the Executive regulation, therefore, exists from the moment in which the reference was created, even if the document has no existing Expression. That this is handled as a Work even if no Expression exists yet is necessary for the consistency of the Legal System and the immodifiability of the enacted laws.

But the legal domain is not the only where such situations exist. For instance, do lost works deserve being described? They have no cataloguing requirement, so according to the authors' claim, they are NOT Works. Are the lost operas by Claudio Monteverdi considerable Works even if we do not have a single fragment of their scores? Or, if we grant special status to Works that had an Expression in the past but not now, how about Works that we do not know for sure whether they actually existed or not? Say the conjectured volume of the Poetics by Aristotle dealing with Comedy? Is it a Work? How about Works that were mentioned as having been written long before they were actually written? For instance, the book "Fantastic Beasts and Where to Find Them" by Newt Scamander (2001), frequently cited in J.K. Rowling's "Harry Potter and the Philosopher's Stone" (1997) much before it was actually written, is it a Work? Amazon sells it. Is the novel "Ship of Theseus", by V.M. Straka, a Work? Amazon sells it. Is the book "It was a Dark and Stormy Night", by Snoopy, a Work? Amazon sells it. These books stretch the very definition of existence: they existed after they were described, and were written exactly because they were described in sufficient detail to make people wish to read them. Can we accept the idea that some need to talk about fictional works as much as others need to talk about fictional people and fictional places, and that the boundaries between fictional and non-fictional entities are sometimes so blurry that is best not to draw firm lines? Is Luther Blisset a real or fictional person? Is Nicolas Bourbaki a real or fictional person? Is Homer a real or fictional person? Who wrote the Bible?

Again, nothing against the authors asserting that they, in their limited and constrained and cozy research context, do not need to deal with Expression-less Works, but the condescending and superior tone with which they deny any need of them whatsoever is overly naive and unacceptable.

And speaking of tones, how condescending and disdainful is, in context, the sentence "we do not attempt at representing marketing or data about pop music bands." that seems to collapse into a single all-encompassing category, "pop", every non-classical style: blues, jazz, country, rock, rap, hip hop, etc., as well as, I suspect, non-western music of all sorts?

And finally, I tend to disagree even with the content of a citation reported by the authors, "Riley argues indeed that FRBR Work is not capable of making sense of improvisations or musical traditions for which there is not a clear composition event leading to a work that is eventually realized in a performance." Nothing to blame on the authors of the paper, but here, too, I suspect the author of this citation has spent little time in pondering how FRBR could, in fact, make a lot of sense of improvisations and non-composed music. Consider on the one hand "Goodbye Pork Pie Hat", a Work by Charles Mingus that has become a standard played (every time in a different way through improvisations) by innumerable jazz artists: all these interpretations are covers of the same piece, and therefore we have one Work with many Expressions. On the other hand, consider "Asbury Park", an improvisation by King Crimson in Asbury Park, NY, June 28 1974, as recorded and available on the live record "USA" (1975). In those years, King Crimson had at least one major improvisation on every concert, every time different in more profound senses than the individual variation over the same themes and correctly reported with a different titles in all recordings and bootlegs existing of that period. For this reasons, these improvisations are not variants of the same Work, but Works on their own: one Works with one Expression. No sweat.

I will not further indulge in this discussion, even if there would be other matters of discussion. For instance, again, the discussion on Aggregation Works reflect the fact that the authors believe belonging to an anthology to be a Work-related property instead of an Expression-related one, which again to me is a fundamental misunderstanding of the model as a whole. But, alas, these are minor problems when compared to the previously mentioned ones, and I will not insist further.

Overall the paper is well written and interesting, and the purpose of the research is noteworthy and meritorious, and I throughly enjoyed reading it and debating it. Nonetheless, at the moment, I believe there are simply too many problems, at the very least in persuading the reader of the correctness of their approach, for this paper to be acceptable in the current form.

I strongly suggest a much humbler approach, where their conceptual model, as they use it for the description and characterization of the scores of their dataset, is described not in opposition to, but in the proper relationship with FRBR(oo), with the correct connections, in which items are items, manifestations are manifestations, expressions are expressions, and works are works, and where the authors simply and plain state that their needs are lesser than the whole of FRBR(oo), and therefore they have carved their model as a smaller one, yet fully compatible with FRBR(oo) itself.

Review #2
Anonymous submitted on 08/Mar/2020
Suggestion:
Reject
Review Comment:

This paper propose a Semantic Web based knowledge and data representation of bibliographical data.
The authors present a solid state of the art of musical knowledge representation and data management.
One of the goal of these works is publishing bibliographical data about music in the Linked Open Data Cloud and make them accessible and reusable.
Nevertheless, the authors point out the limites of the current models of description of bibliographical knowledge and data to which they propose a modeling solution.
The main limitation identified by the authors is the ambiguity in the descritpion of the concepts "Work" and "Expression" of the music in the book Functional Requirements
For Bibliographic Records (FRBR) and its object oriented version (FRBRoo).

After the presentation of the modeling of the "Work" concept in the book FRBR, the authors explained the ambiguity between this concept and "Expression" concept,
especially at the level of the order of instantiation between a "Work" instance and an "Expression" instance.
Indeed, authors state that works are created only a posteriori by abstracting from, e.g., expressions for cataloging purposes,
and differently, the definition of Expression assumes that works are not necessarily realized suggesting that works are not created
by abstracting from expressions and could even preexist expressions and this view in turn now takes them from a problem of data modeling to one of the genesis of
artistic work. So according to the authors, The idea that works precede their expressions is a Romantic one indeed, aligning as it does with
ideas of genius, inspiration, and the Ideal Work.

After presenting this ambiguity, the authors proposed a fine grained taxonomy of the concept "Work" with the concept "documentary Work" as its root.
The authors also presented the OWL definition of the classes of the taxonomy and the OWL definition of the instance examples. They also show how instantiate
"documentary Work" from instance of "Expression".

My main remarks are following :

1 - It would be useful to illustrate with a concrete and practical example the ambiguity of the concept "work" as defined in (FRBR), this will help user to understand.

2 - Lack of explanation and illustration of the interactions between "Expression" and "Document Work", and yet the problem addressed was precisely the interaction between "Work" and "Expression".

3 - It would be useful to make a comparison between instance definition of "Document Work" and "Work" from the same "Expression" in other to show the limit of "Work" modelisation
With only example of instance definition of "Document Work" it is very difficult to appreciate the contribution of this concept.

4 - A clear and formal presentation of retrieving "Document Work" instance from "Expression instance".

5 - Since the authors show that there are multiple musical ontologies for a large spectrum of application, it would be useful to get an evaluation of one of them
using the propose Work taxinomy instead of the one depicted in (FRBR). This evaluation will allow us to get the big picture about :
i) The ambiguity of the concept "work".
ii) The useful and impact of the concept "Document Work" of the contribution.
iii) The added value of the concept "Document Work" in the use of bibliographical ontologies instead of "Work".

Review #3
By Andrea Giovanni Nuzzolese submitted on 18/Mar/2020
Suggestion:
Major Revision
Review Comment:

The paper presents an ontology for representing bibliographical musical data. The ontology results from a critical analysis of FRBR and FRBRoo that points out what are the critical issues arising from FRBR/FRBRoo. Namely, the authors focus on providing a solution for clarifying the intended meaning of Work, which has been widely debated and interpreted in different flowers in literature so far. Additionally, the authors fix the ambiguities of the classes Complex Work and Container Work as defined in FRBRoo. With regard to the latter, the authors adopt a solution that allows to distinguish between (i) compound works whose elements (e.g. movements or sections) form self-contained wholes (e.g. symphonies or masses) and (ii) compound works whose elements are aggregated according to some unifying criteria (e.g. anthologies). Although the authors do not explicitly debate on that, this solution has been discussed in number of works at the state of the art. As a matter of fact, a number of best ontology design practices (a.k.a. ontology design patterns) models mereological relations. Examples are [1], [2], [3], which distinguishes between parts and proper parts, and [4], which models objects that can be simple or aggregated.

==== Overall comments ====
The paper is well written and structured in all its parts.
The paper discusses the problem of modelling bibliographical musical data from a fair perspective, which, in my opinion is worth to be analysed in an ontology paper.
Hence, the paper is on topic with respect to SWJ.

=== Strengths ===
The initial analysis on the issues and ambiguities affecting the interpretation and use of the classes Work, Complex Work, and Container Work of FRBR/FRBRoo is sound. This analysis is supported by references to state of the art works that clarify the nature of the issue or ambiguity in the use of those concepts.
Section 3 provides a good overview on existing ontology for representing musical knowledge with different level of granularity and different perspectives.
This overview is carried out by fairly providing comparisons between existing ontology solutions and that presented by the authors.
Nevertheless, the paper shows some weaknesses that, in my opinion, should be addressed in order to strengthen the paper quality.

=== Weaknesses ===
Those weaknesses are:

+++ Lack of clear ontological requirements and methodology +++
The paper suddenly presents the ontology solutions without introducing clear requirements to be used for both designing and evaluating the ontology. In fact, although the authors largely debate on the weaknesses and ambiguities (cf. Section 2) associated with the classes of FRBR and FRBRoo they want to borrow, they never point out what are the ontological requirements they want to address. Many alternatives ways exist at the state of the art for representing ontological requirements. Competency Questions [4] are one of them. Accordingly, it is not clear what is the methodology adopted for designing the ontology.

+++ Mismatched axioms +++
There is a mismatching between the classes as they are presented in the paper and those available in the ontology available on line. More specifically, some restrictions are defined in the ontology and not presented in the paper.

+++ No plan for maintainability +++
The authors never discuss on the strategies and plans meant for the maintainability of their ontology. For instance, there is no rationale for defining (permanent) identifiers for classes and properties or there is no suggestion about how to deal with possible future changes.

+++ Lack of alignments +++
The ontology is not aligned with any OWL version of FRBR and FRBRoo. This is surprising and counterintuitive.

+++ Evaluation +++
The SPARQL queries presented in Section 5 should provide an indication about the effectiveness of the ontology in capturing the specific domain. However, it is not clear how those queries are derived as no requirement has been introduced.

1. http://ontologydesignpatterns.org/wiki/Submissions:PartOf
2. http://ontologydesignpatterns.org/wiki/Submissions:CollectionEntity
3. http://ontologydesignpatterns.org/wiki/Submissions:Componency
4. Grüninger, Michael, and Mark S. Fox. "The role of competency questions in enterprise engineering." Benchmarking—Theory and practice. Springer, Boston, MA, 1995. 22-31.