The Story of a Soggetto

CRIM@Tours

June 2022

1

1. Encoding Music: From Source to Screen

​

2. Structured Data: For Humans and Machines

​

3. Ontologies: The Structure of Knowledge

​

4. Connecting Knowledge and Data

2

1. Encoding Music: From Source to Screen

​

3

Source and Modern Edition (Sibelius)

4

Sibelius > Music Encoding Initiative > Verovio

5

View MEI File

MEI: One Bar, with Editorial Accidental

6

View MEI File

MEI + EMA (Enhancing Music Addressability)

7

EMA Reference= filename + 8-9/1,1/@all,@all

measures/staves/beats

Read the full MEI specification here

MEI, EMA, Verovio, MEICO

8

MEI score (file.mei)

file.mei/8-9/1,1/@all,@all

EMA selection

To processor

Returns highlighted selection

Retrieves selection

Returns cut selection

2. Structured Data: For Humans and Machines

​

​

9

CRIM Relationships for the Soggetto

10

CRIM analysts have recorded over 500 Relationships between Cadéac’s chanson and the two Masses based on it by Jean Guyon and Nicolas Gombert!

​

Many of these involve the opening pair of soggetti!

​

Examples:

​

​

Structuring and Storing Data: Defining Object Types

​

11

Django objects for persons, works, observations, etc as a relational database.

​

​

​

​

Data saved as JSON = Java Script Object Notation

define CRIM_Observation_Object� has: ID (integer)� links to: observer (CRIM_Person_Object)� links to: piece (CRIM_Piece_Object)

has: Musical_Type� has: EMA expression (string)� has: remarks (string)� has: date of creation (auto-generated)

has details (voices, features

12

Accessing the Data: Django Views

A Django view translates the data into a consumable format, following�any links that are needed to show the user what is desired.

Two renderers:

^{Human-readable: Machine-readable:}

​

​

​

​

​

13

Adding to Django: Views Work Both Ways

​

Webform to JSON

14

Lists and Filters

Django templates present lists of data for the user (by composer, piece, etc)

​

15

Advanced Search via a Python Application

Works like the index at the back of a book!

Reads JSON fields and returns filtered results

​

16

Activity? Hands on with CRIM metdata?

We how have a NB that interacts with CRIM Django metadata.

It reads the lists of persons, relationships, observations, etc and allows fairly simple interaction with the various fields. NOTE THAT VIAF data are in Remarks, not in a regular field!

It could be the focus of some activity, either during the Story of a Soggetto (we search for things relating to this piece and what people say about it)

​

We

17

Graph of Soggetti?

This is a placeholder. DRB is planning a graph of soggetti that will focus on our melody

18

3. Ontologies: The Structure of Knowledge

19

What is a book?

20

"Crime and Punishment by Fyodor Dostoevsky is an heavy book"

What is a book?

The material (digital) item we buy and stock in our homes?

What happens if our book loses a page? Is it always the same book?

The text in a material (digital) item?

What happens when a text is translated? Is it always the same book?

What happens if two persons produce texts with exactly the same words and structure (cf. Pierre Menard in Borges' tale)? How many texts do we have?

The "thing" told? The "content" (meaning) of a text?

… Others?

​

21

What is a musical work?

22

How can we answer these questions?

• "[…] not by looking deeply into the world, but rather by way of conceptual analysis" (Amie Thomasson, (2009). Answerable and unanswerable questions, In Metametaphysics: New Essays on the Foundations of Ontology. Oxford University Press)�
• The way in which we use terms can depend on different sorts of conventions possibly influenced by culture, society, language, and cognition�
• We need to make explicit terms' meanings in order for people (and machines) to communicate and interact!

23

Applied Ontology

An emerging interdisciplinary area building on:

• Philosophy, cognitive science, linguistics, and formal logic with the purpose of understanding, clarifying, making explicit, and communicating people's assumptions about the world�
• This orientation towards helping people understanding each other distinguishes applied ontology from philosophical ontology, and motivates its unavoidable interdisciplinary nature

​

​

24

What is an ontology? (computer science)

An ontology is a model expressing the intended meaning of a vocabulary in a formal (and possibly machine-readable) form…

… in terms of categories and relations describing a domain of discourse (e.g., music, musicology, industrial engineering, biology, etc.), e.g., (informally):

• Musical performance is an event taking place at a certain time and place, having musicians performing a musical composition
• A composer is a person who has created at least one musical composition

25

What is an ontology? (example)

UML Class Diagram notation

Authorial parts!!

See OMAC ontology in OWL

Why ontologies?

Every organization, each computer application adopts a specific vocabulary to talk about a domain of discourse, namely, to organize data

Nowadays, we wish to:

• Share, compare, and possibly even integrate multiple datasets

An ontology is a sort of esperanto for both people and machines

​

27

The Semantic Web

28

Ontological precision

29

Credit to:

Nicola Guarino (CNR)

Intuition about representing data in RDF

RDF triples: Subject Predicate Object, e.g.,

• RF firstName "Richard"
• RF familyName "Freedman"
• RF academicQualification musicology
• RF projectManager CRIMproject
• RF professorAt HaverfordCollege
• RF residentIn Haverford�
• HaverfordCollege locatedIn Haverford
• Haverford locatedIn Philadelphia
• Philadelphia locatedIn USA

30

Do not underestimate the task of ontology-based data modeling: Crime and Punishment in Russian, Italian, and French: how many books do you have in your RDF triples?

RDF (data) graph about Richard Freedman

31

RDF (data) graph about authorial structure of a mass

The mass

The composer

The authorial parts

See RDF file (Turtle syntax)

RDF (data) graph with Chanson and Two derived Masses

33

In CRIM vocabulary, having model for a musical piece means to derive from another musical piece

The Soggetto and Data Associated with it

See here for some details associated with our chanson.

• Composer(s) (there are in fact multiple ascriptions for Je suis desheritée!)
• Editions
• Recordings
• Performances
• Publishers
• Editors
• Genre
• Date(s)
• Performing Forces
• Related Works
• Observations and Relationships

34

Use this tool to create your triples in a (pseudo) RDF syntax.

​

Add your ideas at https://bit.ly/CRIM-RDF

​

(credit to colleagues at the University of Basel, CH)

Intuitions about ontology modeling

UML Class Diagram notation

Authorial parts!!

Each musical work part is authorial part of at most 1 musical work. Hence, e.g., Kyrie-MJSD can NOT be section of two different missas.

Note that:

This restriction can be relaxed (imposing a more flexible cardinality restriction)

See OMAC ontology in OWL

Automatic reasoning

Data + Ontology + Inference Rules that a machine can use to logically infer information that is not explicitly stated.

For instance:

• Since Haverford College is located in Haverford, Haverford is located in Pennsylvania, and Pennsylvania is located in USA, we can infer that Haverford College is located in Pennsylvania and USA, too
• Since Richard is resident in Haverford, and Haverford is located in … (as above), we can infer that Richard is resident in Pennsylvania and USA, too
• Since the Missa Je suis désheritée has section Kyrie-MSJD, and Kyrie-MSJD has subsection Christe-MSJD, we can infer that the Missa has subsection Christe-MSJD, too

36

Ontologies and the Semantic Web for music

Increasing attention, several initiatives and projects:

• Doremus (BNF, Philharmonie de Paris, Radio France)
• Polifonia (EU Project)
• Trompa (EU project)
• Music Ontology
• ….

See MusoW - Musical Data on the Web

​

37

Who wrote Je suis desheritée?

38

Attaingnant 1534: “Lupus”

Attaingnant 1540: Cadéac

Who wrote Je suis desheritée?

39

Note that:

• The graph just says that both Cadéac and Lupus composed Je suis desheritée

​

But we want to say more:

• Lupus has been assigned as the composer of Je suis desheritée by Pierre Attaingnant in a source of 1534
• Cadéac has been assigned as the composer of Je suis desheritée by Pierre Attaingnant in a source of 1540

Who wrote Je suis desheritée?

40

We cover an explicit treatment of scholarly claims (opinions). These are not necessarily true wrt reality.

A Claim about Authorship

Who wrote Je suis desheritée?

41

Multiple claims could be incompatible with each other.

Which one to trust more? This is not a question for ontology !! This is for history and musicology

Conflicting claims about Authorship

Who was Lupus?

But who did Attaingnant think was Lupus? There are many composers with similar names in the period, including Didier Lupi, Johannes Lupi, and several others. An entire pack of Wolves (“lupi”)!

​

According to musicologist Bonnie Blackburn (B.J. Blackburn: The Lupus Problem (diss., U. of Chicago, 1970), Lupus is probably identifiable with Lupus Hellinck, a singer and composer active in Brugges, Ferrara, and Rome during the first decades of the sixteenth century.

42

Who was Lupus … as claimed by Bonnie Blackburn?

​

43

A Claim about Identity

A CRIM Relationship

​

See data here: https://crimproject.org/relationships/6/

According to David Fiala, there is a similarity relation – of type Mechanical Transformation – between Baisez moy (by Josquin Des Prés) and the Sanctus section of the Missa Baisez Moy (by Mathurin Forestier)

The similarity relation is characterized (by David Fiala) in terms of various properties (defined in the CRIM Vocabulary)

A Claim about Similarity

A Claim About Similarity in RDF

Some of the CRIM relations

Based on OMAC ontology:

​

• Identity
• Date
• Similarity
• Value
• Structure

Some conclusions

Ontologies (as Semantic Web models) can be used useful for:

• Organizing data to make data semantics accessible to both humans and machines
• Navigating through data by exploring relations
• Publishing data, e.g., through Web resources
• Integrating multiple datasets

Various research works about ontologies for performing arts

• Need for modeling approaches making explicit scholarly claims in machine processable terms (→ digital scholarly criticism)

46

Thank you!

47

Some suggestions

Before developing your models:

• Focus on what you wish to say, i.e., what your model needs to express: What is my vocabulary (categories, relations)?
• Focus on the meaning of the terms in your vocabulary (conceptual analysis): What does this-and-that mean (in my community)?
• Take a critical stance. If you consider reusing existing ontologies (gold practice), BE SURE that they tell what YOU wish to tell!
• If you do not feel comfortable enough with conceptual analysis, applied ontology, formal representations… keep on and read some books ;)

AI is good… but Human Intelligence, too!

48