Interactive Generation of Musical Corpora for Piano Education:

Opportunities and Open Challenges

Filippo Carnovalini

1 a

, Antonio Rod

and Geraint A. Wiggins

2,3

Centro di Sonologia Computazionale, University of Padova, via Gradenigo 6, Padova, Italy

Vrije Universiteit Brussel, Pleinlaan 9, 1050 Brussel, Belgium

Queen Mary University of London, Mile End Road, London E1 4NS, U.K.

Keywords:

Music Education, Integrated Teaching Systems, Music Generation, Computational Creativity.

Abstract:

Learning to play a musical instrument such as a piano requires many hours of exercises, generally taken from

a “method” book. These books are collections of progressive exercises intended to teach speciﬁc techniques

and address the commonest mistakes and difﬁculties that players face while learning. One downside of these

books is that the exercises are not personalized to the students and thus cannot address speciﬁc difﬁculties and

characteristics of each learner. Given the many recent advances in the ﬁeld of music generation, we propose

that it should be possible to generate exercises automatically to form a personalized method for each student.

The teacher would describe the characteristics of the student and their strengths and weaknesses to a software

system, as well as the teaching goals that should be covered in the generated exercises, and the system would

create exercises that are speciﬁc to the needs of the student and the concerns of the teacher, allowing for a

more effective and engaging learning experience.

In this paper, we describe a project trying to design such a system, stating research questions, describing

the tentative methodology, and outlining its potential impact for both research in music generation and in

computer-supported education.

1 INTRODUCTION

Computational Creativity is, as deﬁned by the

homonymous Association

, “the art, science, phi-

losophy, and engineering of computational systems

which, by taking on particular responsibilities, exhibit

behaviors that unbiased observers would deem to be

creative.” Having worked both on musical applica-

tions of Computational Creativity and on Computer-

Supported Education in the past, the authors of this

position paper wish to explore how the two disciplines

can ﬁt together. We believe that Computational Cre-

ativity, and more speciﬁcally Music Generation, can

open novel avenues for music education, and that the

exploration of these new applications can also give

further insights into creativity and its computational

formalization.

In the ﬁrst part of this paper, we will describe

some open problems of these two disciplines, search-

ing for those contact points that we believe can lead

to fascinating hybridization of approaches. Then,

https://orcid.org/0000-0002-2996-9486

https://computationalcreativity.net

we will focus on some speciﬁc research questions

that we wish to address in future work, and intro-

duce the project CALIOPE (Co-creativity And Learn-

ing: Interactive Opus generation for Piano Educa-

tion), which was recently awarded funding by the Eu-

ropean Union under a Marie Skłodowska-Curie Ac-

tions Postdoctoral Fellowship and is set to start in

early 2024. This project will speciﬁcally address

these research questions, hopefully giving novel in-

sights into Computer-Supported Music Education and

Music Generation, by creating a prototype system for

the automatic generation of piano exercises tailored

around a description of the learner’s characteristics

and skill level.

2 STATE OF THE ART

2.1 Music Generation

While the interest in AI-generated music is growing

recently, the task of algorithmically creating music is

not novel and has examples that even predate com-

Carnovalini, F., Rodà, A. and Wiggins, G.

Interactive Generation of Musical Corpora for Piano Education: Opportunities and Open Challenges.

DOI: 10.5220/0012059200003470

In Proceedings of the 15th International Conference on Computer Supported Education (CSEDU 2023) - Volume 1, pages 407-412

ISBN: 978-989-758-641-5; ISSN: 2184-5026

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

407

puters (Nierhaus, 2009). Among some of the more

common approaches to music generation are the use

of statistical methods (e.g., Markov Chains) to pre-

dict suitable notes to continue a musical sequence, the

use of formal grammars to create sequences, genetic

algorithms, and rule-based systems paired with op-

timization algorithms (Carnovalini and Rod

a, 2020).

However, in recent years the method that has be-

come most common for the generation of musical se-

quences is the use of Neural Networks, commonly

applying Deep Learning approaches such as Genera-

tive Adversarial Networks, Convolutional Networks,

Autoencoders, and Long Short-Term Memory Net-

works (LSTM). The latter have proven to be par-

ticularly suitable for the generation of sequences in

time, which can well represent melodies (Briot and

Pachet, 2020). Although many advancements have

been made in this ﬁeld, there are still a variety of

open challenges that need to be addressed. One of

the most notable difﬁculties of music generation al-

gorithms is the lack of long-term structure: gener-

ated pieces are credible for a couple of bars but seem

to “wander off” as the piece progresses, lacking the

formal aspects of repetition and reuse that character-

ize human-composed melodies (Carnovalini, 2019).

Interestingly, even if Machine Learning approaches

have become more common for this task, these sys-

tems offer little control over the generated results

(Briot and Pachet, 2020) making it harder to solve

this problem. Some researchers have started design-

ing ad hoc architectures to obtain well-structured re-

sults (Roberts et al., 2018), while others have em-

bedded multiple pre-existing architectures to delegate

structural aspects to a specialized part of the system

(Cho et al., 2014). Other researchers have directly

addressed the problem of long-term structure. Mor-

pheus (Herremans and Chew, 2017) is another EU-

funded project that leveraged a specialized mathe-

matical representation of harmonic tension to recre-

ate the “tension proﬁle” learnt from existing pieces in

newly generated pieces. Hierarchical representations

of melodic content can also be used to describe reuse

of melodic content and obtain multi-layered descrip-

tions of structure in music pieces (Carnovalini et al.,

2021b).

We believe that the way forward is to expand on

hierarchical representations of melodic content, as

these representations can capture cognitive aspects of

musical structure (Temperley, 2011). Applying such

structures to the context of musical exercises could

also open new directions into the description of stylis-

tic coherence and reuse of melodic content between

multiple pieces, considering long-term structure at a

corpus/book level rather than at the song/piece level.

2.2 Computer Assisted Music

Education

Music Education in a traditional Western context in-

cludes a variety of aspects, all of which can beneﬁt

from the use of computational systems to some ex-

tent. It is possible to divide these aspects into four

main categories: fundamentals of music (music the-

ory), performance skills, music analysis, and compo-

sition (Brandao et al., 1999). In this paper, we wish to

focus on performance skills, i.e., those skills related to

playing one instrument (the piano). Research shows

that the use of a Computer-Assisted Instruction (CAI)

framework can effectively increase the effectiveness

of piano lessons compared to a control group (Kaleli,

2020). In this framework, the computational system

does not substitute the role of the teacher, but rather

complements the teaching by offering a learning envi-

ronment (Abdullah and Mustafa, 2019). For example,

a CAI system can embed score following to help a

student visualize (through a display or through aug-

mented reality) what is being played and the mistakes

they make (Smoliar et al., 1995; Rigby et al., 2020).

Another approach is to use an Integrated Teaching

System (ITS), that is, a system that has greater con-

trol over the teaching activity by proactively helping

the teacher by leveraging knowledge about the subject

being taught, the curriculum, and the student. In this

context, there are systems that can select exercises

from a given corpus based on the needs of the stu-

dent and the teacher’s concerns (Wiggins and Trewin,

2000; S

ebastien et al., 2012). To the best of our

knowledge, there is no system that generates novel

piano exercises to propose to students in a similar

fashion. One similar project is the one by Takidaira

et al (Takidaira et al., 2022), that generates rhythmic

scores for the popular rhythm game Taiko no Tatsujin

to improve the player’s ability in the game by analyz-

ing their game logs. While this is not directly meant

for musical education, performance in a rhythm game

is strongly related to musical skills, and thus this can

be considered a music education tool. However, that

system only focuses on rhythmic aspects, making the

system quite limited compared to the one we pro-

pose. Other researchers dealing with music gener-

ation have considered applications in the context of

learning an instrument, and there is some research to-

wards controlling the difﬁculty of a generated music

piece. Ariga et al (Ariga et al., 2017) proposed a

system to generate guitar solos using ﬁngering as a

means of controlling difﬁculty. Nakamura and Yoshii

(Nakamura and Yoshii, 2018) describe a system that

generated piano reduction of ensemble scores, allow-

ing different difﬁculty settings based on tempo and

CSME 2023 - 4th International Special Session on Computer Supported Music Education

408

ﬁngering, and similar techniques have been applied

to score difﬁculty analysis (Ramoneda et al., 2022).

However, none of these systems really try to build an

ITS that uses music generation for the goal of piano

education.

The current level of customization that is possible

with ITS is limited by the selected corpora and there-

fore cannot offer a fully personalized learning experi-

ence. We posit that the use of computational creativity

can improve the personalized learning experience by

creating a speciﬁc music generation system that takes

into account the characteristics of the student.

2.3 Research Questions

Wishing to explore further levels of personalization

for musical education, while also deepening our un-

derstanding of music cognition and representation for

Computational Creativity, we formulate the following

research question:

Q1: How can Computational Creativity be applied to

improve the quality and personalization of musi-

cal exercises used in a training course?

Two sub-questions arise from this:

Q2: How can hierarchical knowledge representation

improve the longer-term structure of computation-

ally created sequences of exercises?

Q3: How can a model of the learner and the educa-

tional context improve the computational creation

of sequences of exercises?

We decided to address these questions through the

development of a prototype system for teaching pi-

ano. The objective would be to create a system ca-

pable of generating musical exercises in a progressive

manner, considering the playing skills of the user as

input. This goal can be further divided into three sub-

objectives:

SO1: Create a system for the generation of a set of

music pieces with progressive difﬁculty.

SO2: Design a knowledge representation to describe

the user’s weaknesses, strengths, learning needs,

and personality (Learner Model).

SO3: Integrate the Learner Model into the generation

system, so that the generated exercises are person-

alized to the student.

The next section will describe in more detail the

design choices of the project and the features we de-

sire from the ﬁnal prototype, as well as the methodol-

ogy we wish to follow both for the implementation of

the prototype and for ﬁnding answers to the research

questions outlined above.

3 PROJECT CALIOPE

In this section, we wish to introduce project

CALIOPE (Co-creativity And Learning: Interactive

Opus generation for Piano Education), which stems

from the desire to address how Computational Cre-

ativity can help improve personalization in music ed-

ucation. This project was recently awarded a Postdoc-

toral Fellowship under the Marie Skłodowska-Curie

Actions funding scheme. Although the project has

not started yet, this section will introduce the ideas

and design choices that drove the formulation of this

project, hoping to receive comments and suggestions

from the scientiﬁc community before the project’s of-

ﬁcial beginning.

3.1 Summary

When learning to play a musical instrument in a clas-

sical western setting, such as conservatory or music

school, teachers employ exercise books meant to in-

crease the difﬁculty of the provided studies progres-

sively. Such a book is known as a “method”, and it is

easy to ﬁnd numerous methods for any orchestral in-

strument. One of the major drawbacks of these books

is that they do not provide a personalized learning ex-

perience as they do not focus on the speciﬁc needs

of each student, but instead provide general exercises

that can be useful to most. It is the role of the teacher

to select exercises that are most ﬁt for their students,

potentially selecting exercises from different methods

or even creating ad hoc exercises for each student.

This activity requires abundant time and effort from

teachers, who often resort to simply following the

most popular methods instead of personalizing edu-

cation, limiting the effectiveness of music education,

and requiring more effort from students.

Given the considerable recent advancements in

music generation, we propose that it should be pos-

sible to employ computers to automatically generate

personalized methods that are tailored to the needs

of each student. Therefore, the main output of the

project will be a prototype Integrated Teaching Sys-

tem for piano education. A music teacher would use

this system to generate exercises that are tailored to

their students, by giving the system information about

each student via specialized knowledge representa-

tions to describe the learner, called the Learner Model

(Bull, 2020). The model describes the strengths,

learning needs, and personality of the student. After

the initial setup of the learner model, the prototype

will give the teacher some starting exercises for the

student. The student practices using the generated ex-

ercises and the teacher evaluates their performance.

Interactive Generation of Musical Corpora for Piano Education: Opportunities and Open Challenges

409

Based on that, the teacher can describe to the system

what kind of difﬁculties the student found and what

they should focus on going forward, and what is the

goal of the following practice session (i.e., what is

the Teaching Concern (Wiggins and Trewin, 2000)).

The system can use this information to generate fur-

ther exercises, possibly also including annotations on

things that the learner should pay attention to while

exercising, allowing the learning cycle to progress.

In this way, the system would help provide a per-

sonalized exercise plan based on the characteristics of

the student, addressing the weaknesses of the student

while maintaining a satisfactory learning curve since

the generation would consider their current level of

expertise.

3.2 Methodology

To create the system described above and answer the

research questions described in the previous section,

the following methodological steps will be employed:

(i) Creation of the dataset. The AI system will re-

quire a dataset of relevant piano methods. Many

exist, but at least 10 methods must be selected

(estimating about 50 exercises per method for

a total of 500 exercises). Methods in the pub-

lic domain, available digitally, and that respect

gender balance will be preferred. Syllabi for pi-

ano courses at music education institutes will be

used to ﬁnd well-established methods.

(ii) Model of Difﬁculty. A detailed study of the

literature on automatic difﬁculty analysis (Ra-

moneda et al., 2022) will serve as a basis for

the deﬁnition of a model of difﬁculty. Using a

top-down approach to the matter, to deﬁne con-

straints that describe difﬁculty.

(iii) Generation of Progressive Exercises. Hierar-

chical representations of structure, jointly with

machine learning (different algorithms will be

tested, including LSTM networks), will be

leveraged to generate musical pieces. The dif-

ﬁculty model will be embedded in the genera-

tion system as a self-reﬂection module that in-

ﬂuences the output.

(iv) Evaluation (Difﬁculty Generation). A meta-

evaluation procedure (Carnovalini et al., 2021a)

will be used to evaluate the difﬁculty control

system. Expert-based evaluation (e.g., CAT

(Amabile, 1983), SPECS (Jordanous, 2012))

will be used to assess the quality of the pro-

duced exercises.

(v) Learner Model Design. A collaborative design

(Chiu, 2002) approach will be used, creating a

focus group of piano teachers that will help me

describe the characteristics of the student in the

model, as well as the teaching concerns in piano

education.

(vi) Evaluation (Learner Model). A focus group

will be used again to evaluate how well the de-

signed model captures the requirements.

(vii) Personalization. The difﬁculty model will be

expanded to consider the learner model.

(viii) Final Evaluation. The same as point iv, but also

using a test set covering all the limit use cases

of the learner model will be used, and the ex-

pert evaluation will consider the different ap-

plied settings.

The methodology is dictated by the diverse objec-

tives of the project. Computational creation of mu-

sical content requires the use of musical information,

which can be provided either via data in a bottom-up

manner or via description of rules and constraints in a

top-down manner. For the generation of musical ma-

terial, the bottom-up approach is chosen as literature

shows that machine learning is effective in describing

musical sequences, with added top-down information

given by the hierarchical model to ensure long-term

coherence. For the description of difﬁculty and of

the learner model a top-down approach is used be-

cause machine learning in this context would require

higher amount of musical data with difﬁculty annota-

tions and might fail to capture the multidimensional

concept of difﬁculty. To capture the real-life teach-

ing requirements that piano teachers meet, collabora-

tive design is used for SO2. This will help maintain a

human-centric design and avoid bias introduced by AI

in the most sensitive part of the project, which is the

description of the learner model. Evaluation follows

the principles established by the literature for compu-

tational creativity and music generation, requiring ex-

perts to gather insights on the capabilities of the sys-

tem. The use of limit use cases for the ﬁnal prototype

serves as a proxy for the evaluation of the impact of

this teaching system using actual students, which can

only be considered in a longer-term study when the

prototype created with this project has already been

thoroughly tested in an ofﬂine setting.

3.3 Relevance and Open Directions

Being a project that stems from research ques-

tions coming from both Computational Creativity and

Computer-Supported Education, the potential impact

of the project is twofold.

From the viewpoint of music generation, this

project offers a novel approach to the problem of

CSME 2023 - 4th International Special Session on Computer Supported Music Education

410

long-term structure: multi-piece generation. By con-

sidering the entire method as the generated output,

this project will need to generate a large amount

of musical content that must be coherent and well-

structured, both considering the usual criteria for mu-

sical well-formedness within each piece and consid-

ering higher structures joining the different exercises,

maintaining a coherent style throughout. The fact that

exercises are distinct music pieces gives some leeway,

but the additional constraints regarding progressive

difﬁculty and educational value make this an intrigu-

ing use case. Under the broader perspective of non-

strictly musical Computational Creativity, the project

offers a novel take on the exploration of a concep-

tual space (Boden, 2004). The Learner Model serves

to deﬁne personalized conceptual spaces that are ex-

plored by the generation algorithm as per SO3: the

space of the possible musical pieces is constrained

and transformed dynamically by the Learner model,

and the generation of novel pieces both explores the

deﬁned space by adapting the difﬁculty of pieces, and

by transforming the space of useful exercises by up-

dating the Learner model. This approach also deals

with the problem of Explainability in AI: by having

the system generate music that is speciﬁcally tailored

to address a speciﬁc teaching concern, it is possible to

assess how well the system understands the given task

and it would be possible to leverage the knowledge in

the system to describe what makes each exercise apt

for the task at hand.

From the viewpoint of Computer-Supported Ed-

ucation, this project will offer a case study for the

employment of Artiﬁcial Intelligence in the person-

alization of the educational path. This technology

can offer novel insight on how an Integrated Teach-

ing System can support both the Teacher and the Stu-

dent through the use of the novel human-computer in-

teraction routes offered by Artiﬁcial Intelligence and

Computational Creativity.

Finally, the creation and public release of the

data and technologies related to this project (exercise

dataset, learner model, difﬁculty analyzer) will hope-

fully foster further research both in music generation

and in Computer-Supported Education.

4 CONCLUSIONS

In this position paper we presented project CALIOPE,

by giving an overview of the research questions it will

try to answer, the methodology that we intend to fol-

low to address those questions, and the potential im-

pact of the project. In the latter part of the paper, we

brieﬂy reviewed relevant scientiﬁc literature.

The main goal of the project is to develop a proto-

type for an Integrated Teaching System that will lever-

age Music Generation techniques for the creation of

personalized piano exercises. By doing so, we believe

it will be possible to address some open problems of

music generation and computational creativity, while

advancing the knowledge in Computer-Supported ed-

ucation.

Although the project is still in its preliminary

phase, the authors are open to suggestions and will-

ing to ﬁnd potential collaborations from the scientiﬁc

community. With this paper, our desire was to state

our intents and what we believe could result from this

study, and let the project be known to the relevant

community.

ACKNOWLEDGEMENTS

The ﬁrst author wishes to thank his supervisors for all

the help given in these years and for the sincere inter-

est in the author’s career development.

While this paper describes a project for which fund-

ing was approved by the European Union, the cur-

rent work was not funded by the European Union and

does not necessarily reﬂect the views of the European

Union, and the authors are the sole responsible for the

contents of this article.

This work was partially supported by the project Cre-

ative Recommendations to Avoid Unfair Bottlenecks,

granted by the Department of Information Engineer-

ing, University of Padova.

REFERENCES

Abdullah, Z. D. and Mustafa, K. I. (2019). The Underlying

Factors of Computer Self-Efﬁcacy and the Relation-

ship with Students’ Academic Achievement. Interna-

tional Journal of Research in Education and Science,

5(1):346–354. Publisher: ERIC.

Amabile, T. M. (1983). A Consensual Technique for Cre-

ativity Assessment. In Amabile, T. M., editor, The So-

cial Psychology of Creativity, Springer Series in So-

cial Psychology, pages 37–63. Springer New York,

New York, NY.

Ariga, S., Fukayama, S., and Goto, M. (2017).

Song2Guitar: A Difﬁculty-Aware Arrangement Sys-

tem for Generating Guitar Solo Covers from Poly-

phonic Audio of Popular Music. In ISMIR, pages 568–

574.

Boden, M. A. (2004). The creative mind: Myths and mech-

anisms. Routledge, London, United Kingdom.

Brandao, M., Wiggins, G., and Pain, H. (1999). Computers

in music education. In Proceedings of the AISB’99

Symposium on Musical Creativity, pages 82–88. AISB

Interactive Generation of Musical Corpora for Piano Education: Opportunities and Open Challenges

411

(Society for the Study of Artiﬁcial Intelligence and the

Simulation of . . . .

Briot, J.-P. and Pachet, F. (2020). Deep learning for music

generation: challenges and directions. Neural Com-

puting and Applications, 32(4):981–993.

Bull, S. (2020). There are Open Learner Models

About! IEEE Transactions on Learning Technologies,

13(2):425–448.

Carnovalini, F. (2019). Open Challenges in Musical

Metacreation. In Proceedings of the 5th EAI Inter-

national Conference on Smart Objects and Technolo-

gies for Social Good, pages 124–125, Valencia Spain.

ACM.

Carnovalini, F., Harley, N., Homer, S. T., Rod

a, A., and

Wiggins, G. A. (2021a). Meta-Evaluating Quantita-

tive Internal Evaluation: a Practical Approach for De-

velopers. In Proceedings of the 12th International

Conference on Computational Creativity, page 5,

exico Virtual. Association for Computational Cre-

ativity.

Carnovalini, F. and Rod

a, A. (2020). Computational Cre-

ativity and Music Generation Systems: An Introduc-

tion to the State of the Art. Frontiers in Artiﬁcial In-

telligence, 3:14.

Carnovalini, F., Rod

a, A., Harley, N., Homer, S. T., and

Wiggins, G. A. (2021b). A New Corpus for Compu-

tational Music Research andA Novel Method for Mu-

sical Structure Analysis. In Audio Mostly 2021 (AM

’21), page 4, virtual/Trento Italy. ACM.

Chiu, M.-L. (2002). An organizational view of design com-

munication in design collaboration. Design Studies,

23(2):187–210.

Cho, K., Van Merri

enboer, B., Gulcehre, C., Bahdanau, D.,

Bougares, F., Schwenk, H., and Bengio, Y. (2014).

Learning phrase representations using RNN encoder-

decoder for statistical machine translation. arXiv

preprint arXiv:1406.1078.

Herremans, D. and Chew, E. (2017). MorpheuS: gener-

ating structured music with constrained patterns and

tension. IEEE Transactions on Affective Computing,

10(4):16.

Jordanous, A. (2012). A standardised procedure for evalu-

ating creative systems: Computational creativity eval-

uation based on what it is to be creative. Cognitive

Computation, 4(3):246–279.

Kaleli, Y. S. (2020). The Effect of Computer-Assisted

Instruction on Piano Education: An Experimental

Study with Pre-Service Music Teachers. Interna-

tional Journal of Technology in Education and Sci-

ence, 4(3):235–246. Publisher: ERIC.

Nakamura, E. and Yoshii, K. (2018). Statistical piano re-

duction controlling performance difﬁculty. APSIPA

Transactions on Signal and Information Processing,

Nierhaus, G. (2009). Algorithmic Composition: Paradigms

of Automated Music Generation. Springer Science &

Business Media. Google-Books-ID: jaowAtnXsDQC.

Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X.,

and Miron, M. (2022). Score Difﬁculty Analysis

for Piano Performance Education based on Finger-

ing. In ICASSP 2022 - 2022 IEEE International Con-

ference on Acoustics, Speech and Signal Processing

(ICASSP), pages 201–205, Singapore. IEEE.

Rigby, L., W

unsche, B. C., and Shaw, A. (2020). piARno

- An Augmented Reality Piano Tutor. In 32nd Aus-

tralian Conference on Human-Computer Interaction,

pages 481–491, Sydney NSW Australia. ACM.

Roberts, A., Engel, J., Raffel, C., Hawthorne, C., and Eck,

D. (2018). A hierarchical latent vector model for

learning long-term structure in music. In International

conference on machine learning, pages 4364–4373.

PMLR.

Smoliar, S. W., Waterworth, J. A., and Kellock, P. R. (1995).

pianoFORTE: a system for piano education beyond

notation literacy. In Proceedings of the third ACM

international conference on Multimedia - MULTIME-

DIA ’95, pages 457–465, San Francisco, California,

United States. ACM Press.

ebastien, V., Ralambondrainy, H., S

ebastien, O., and Con-

ruyt, N. (2012). Score analyzer: Automatically de-

termining scores difﬁculty level for instrumental e-

learning. In ISMIR 2012, pages 571–576.

Takidaira, S., Shoji, Y., and D

urst, M. J. (2022). Slow, Re-

peat, Voice Guidance: Automatic Generation of Prac-

tice Charts to Improve Rhythm Action Games. In

Fang, X., editor, HCI in Games, volume 13334, pages

119–138. Springer International Publishing, Cham.

Series Title: Lecture Notes in Computer Science.

Temperley, D. (2011). Composition, Perception, and

Schenkerian Theory. Music Theory Spectrum,

33(2):146–168.

Wiggins, G. A. and Trewin, S. (2000). A System for Con-

cerned Teaching of Musical Aural Skills. In Goos, G.,

Hartmanis, J., van Leeuwen, J., Gauthier, G., Frasson,

C., and VanLehn, K., editors, Intelligent Tutoring Sys-

tems, volume 1839, pages 494–503. Springer Berlin

Heidelberg, Berlin, Heidelberg. Series Title: Lecture

Notes in Computer Science.

CSME 2023 - 4th International Special Session on Computer Supported Music Education

412