The problems associated with the design, implementation and evaluation of an intelligent piano tutor

  • Robert John Boland

    Student thesis: Doctoral Thesis


    The overall goal of the project is to produce experiential information concerning the design, construction and use of an Intelligent Tutoring System (ITS) for educational purposes, and how this information can be represented over a number of levels of abstraction. To generate and collect this information, an Integrated learning Environment (ILE) has been set up containing three components, an adult human music student, a computer Piano Tutor designed and programmed by the author, and a set of instructions, the Practice Script, prepared by the student’s human music teacher. Over the lifecycle of the project, a practical top down approach to ITS design has been taken, beginning with adult education theory and principles, through architectural modelling and software design, to physical implementation of a fully functional Piano Tutor, built as proof of concept. The Piano Tutor is considered as a complex reactive machine, which can adapt to its environment and is represented in the dissertation at five hierarchal levels of abstraction, beginning in Chapter 1 at the top, level V, with a statement of the Overall Goal or system purpose. Level I (the lowest) is presented in Chapter 2, as an introduction and Overview of the complete Physical System, and describes the interactions between the Student, the Piano Tutor and the Practice Script within the ILE. In this overview, the physical components which make up the Tutor are discussed, as are the physical characteristics of the ILE. Level IV is a representation, presented in Chapter 3, of the Educational Plan for the set up and maintenance of the ILE in which an adult music student can practice piano keyboard techniques under the guidance of the Piano Tutor. Fundamental elements of curriculum design, such as the way adults learn, the range of teaching styles a music teacher can have, what could reasonably be expected of an ideal machine tutor, and for an ITS, are discussed. A comprehensive curriculum design model for the Piano Tutor is developed from basic principles and presented. The domain of Music is not discussed as a topic in its own right in this dissertation. However, examples drawn from basic music theory and practice are used as a lingua franca across most abstract level boundaries to illustrate concepts and experiences. In Chapter 4, Level III is represented by an Architectural Design view of the Piano Tutor and its instruction set. The latter takes the form of a Practice Script, prepared by the teacher for her student. The Piano Tutor is presented as a set of virtual modules which make up the adaptive reactive system, closely following classical ITS architecture. The purpose of each module, and its interactions with the control system, are discussed in terms of the total system’s reactions to input from the student, and the nature of the instructions in the Practice Script. The architecture of the Practice Script, which is considered as internal system software, is discussed in terms of five basic elements of musical expression and the music keyboard skills associated with them. These elements are touch, tempo, articulation, phrasing and the use of the pedal. Level II is presented in Chapter 5, and is a representation of the Physical Software System which makes up the Piano Tutor. This representation includes critical window objects, performance handling callback functions, and the various helper modules, such as the display module and the fuzzy logic engine. It also details built-in utilities and V tools for diagnosis and for building the various music resources used by the Tutor, and how these files are constructed and utilised. From time to time, code fragments are used to illustrate various points. Testing of the Piano Tutor in a field setting has been an ongoing process since the first simple prototype was built. The results have been used primarily to validate the design, and in the grounding of concepts and perceptions originating from educational and music-theoretic principles. In Chapter 6, the information and experience gained from the Project are presented in terms of the evaluation strategy developed in that chapter. The raw data from a representative sample of six short test sessions with the current prototype Piano Tutor are presented with a few short observer’s comments to allow readers to draw their own conclusions as to whether the “tutor can help an adult learn”. General and specific problems identified throughout the lifecycle of the project are selected on the grounds of possible usefulness as learning resource materials in a postgraduate Instructional Design course, and are outlined and discussed in Chapter 7. No attempt has been made to directly quantify student learning, but future directions for improvement of the Piano Tutor and some areas of music learning theory, in which it might prove useful as an intelligent research tool, have been identified throughout the dissertation.
    Date of Award2003
    Original languageEnglish
    SupervisorJim Mitchell (Supervisor)

    Cite this