Alan Smaill - Böcker

The research fields of "artificial intelligence and music" and "cognitive musicology" are relative newcomers to the many interdisciplinary groupings based around the centre of AI and cognitive science. They are concerned with the computational study and emulation of human behaviour with respect to music, in many aspects, and with varying degrees of emphasis on psychological plausibility. Recent publications have included work in such diverse areas as rhythm and pitch perception, performance, composition, and formal analysis. Music shares with language the property of giving access to human mental behaviour in a very direct way. As such, it has the potential to be a very useful domain for AI work. Furthermore, in the course of time, AI related work will surely throw light back onto some or all of the fields to which it is applied. Indeed, we are already beginning to feel the benefits of the application of AI techniques to music technology. It is not surprising, therefore, that one of the first areas interest for of musical AI study is that of music education. There are many ways in which an artificial intelligence or cognitive science approach to music education may be applied - for example, to automate tuition, to explain learning processes, to provide metaphors for human computer interaction, and so on. This collection of papers, which is intended to give an impression of both the breadth and depth of the field, originated from a workshop entitled "Music Education: An Artificial Intelligence Approach".

Thisvolumecontainsthefullresearchpapersofthe2nd International Con- renceonMusicandArti?cialIntelligence(ICMAI02),heldinSt. Cecilia'sHall, Edinburgh,UK,12-14September2002. Theconferencewasjointlyorganizedby theFacultyofMusicandtheDivisionofInformatics,UniversityofEdinburgh. Eachofthepapersinthisvolumewasrefereedbyatleastthreereviewersfrom theProgramCommittee. AdditionalpaperswereacceptedasWork-in-Progress reports,publishedseparatelybytheDivisionofInformaticsoftheUniversity ofEdinburgh. Theconferenceprogramalsoincludedround-tablediscussions, electronicmusicconcerts,earlykeyboardintrumentdemonstrations,andvarious socialevents. Thefocusoftheconferencewasontheinterplaybetweenmusicaltheoryand practiceontheonehandandtechniquesandmethodsfromArti?cialIntelligence, includingcomputationalcognitivemodelingofmusicalprocesses,ontheother. Weareespeciallyinterestedinissuesofmusicalrepresentationandanalysis, associatedalgorithms,andtheevaluationofmusicalartefacts. MusicpresentsmanychallengesforAIandInformaticsingeneral. Whilethe analogieswithnaturallanguageandwithvisionareproductive,musicrequires novelsolutionstoitsownrepresentationalandalgorithmicproblems.Forthe musician,workinthisareacontributestowardsmusicologicalstudy,composition, performance,interactivesystems,andsoundsynthesis. Threedistinguishedscholarsinthearea,MiraBalaban,Jean-ClaudeRisset, andAntonioCamurri,agreedtogivetheinvitedkeynotes. Thesetalkscovered di?erentfeaturesofmusicalexperienceanditsrelationtoArti?cialIntelligence, namelymusicalconcepts,composition,andperformance. Morespeci?cally,Jean- ClaudeRisset'stopicwasMusicalCompositionandArti?cialIntelligence:Some Precedents and Prospects,MiraBalaban'swas Structure and Interpretation of Music Concepts - Music from a Computational Perspective,andAntonio- murri'swasComputationalModelsofExpressiveGesture. Thepapersintheconferencedealtwithawiderangeofaspectsofcurrent research,includingstructural,harmonic,andreductionalmusicanalysis,pattern representationanddiscoveryinbothmonophonicandpolyphonicmusic,music perceptionofmelodyandrhythm,therelationbetweenmusicandnaturallan- age,similarityandcategorization,musicandintonation,musicalexpressionand performance,soundprocessing,soundclassi?cations,commercialapplications, andmusicontheweb. Acknowledgements.Theeditorsaregratefulforallthehelpofthosewho madethepublicationofthisvolumepossible,andespeciallyDarrellConklin, VI Organization PeterNelson,EmiliosCambouropoulos,AlisonPease,andMarkSteedman;also, JordanFleming,EwenMaclean,andRobertDow;andtheFacultyofMusicand theDivisionofInformatics,UniversityofEdinburghfor?nancialsupport. July2002 ChristinaAnagnostopoulou MiguelFerrand AlanSmaill OrganizingCommittee AlanSmaill(Chair) ChristinaAnagnostopoulou(Co-chair) MiguelFerrand PeterNelson AlisonPease MarkSteedman ProgramCommittee ?JensArnspang(Alborg) GillianHayes(Edinburgh) MiraBalaban(Ben-Gurion) HenkjanHoning(Nijmegen) MartaOlivettiBelardinelli(Rome) JukkaLouhivuori(Jyvaskyla) RensBod(Amsterdam) TodMachover(MIT-MediaLab) CarolaBoehm(Glasgow) GuerinoMazzola(ZurichandVienna) AmilcarCardoso(Coimbra) StephenMcAdams(IRCAM) EmiliosCambouropoulos(Thessaloniki)EduardoMiranda(SONYCSL-Paris) ElaineChew(USC) PeterNelson(Edinburgh) DarrellConklin(ZymoGenetics) Fran,coisPachet(SONYCSL-Paris) IanCross(Cambridge) StephenTravisPope(SantaBarbara) RogerDannenberg(CMU) AlanSmaill(Edinburgh) PeterDesain(Nijmegen) MarkSteedman(Edinburgh) AnastasiaGeorgaki(Athens) DavidTemperley(Rochester) InvitedSpeakers MiraBalaban(Ben-GurionUniversity,Israel) AntonioCamurri(UniversityofGenoa,Italy) Jean-ClaudeRisset(LaboratoiredeM'echaniqueetd'14September2002. Theconferencewasjointlyorganizedby theFacultyofMusicandtheDivisionofInformatics,UniversityofEdinburgh. Eachofthepapersinthisvolumewasrefereedbyatleastthreereviewersfrom theProgramCommittee. AdditionalpaperswereacceptedasWork-in-Progress reports,publishedseparatelybytheDivisionofInformaticsoftheUniversity ofEdinburgh.Theconferenceprogramalsoincludedround-tablediscussions, electronicmusicconcerts,earlykeyboardintrumentdemonstrations,andvarious socialevents. Thefocusoftheconferencewasontheinterplaybetweenmusicaltheoryand practiceontheonehandandtechniquesandmethodsfromArti?cialIntelligence, includingcomputationalcognitivemodelingofmusicalprocesses,ontheother. Weareespeciallyinterestedinissuesofmusicalrepresentationandanalysis, associatedalgorithms,andtheevaluationofmusicalartefacts. MusicpresentsmanychallengesforAIandInformaticsingeneral. Whilethe analogieswithnaturallanguageandwithvisionareproductive,musicrequires novelsolutionstoitsownrepresentationalandalgorithmicproblems. Forthe musician,workinthisareacontributestowardsmusicologicalstudy,composition, performance,interactivesystems,andsoundsynthesis. Threedistinguishedscholarsinthearea,MiraBalaban,Jean-ClaudeRisset, andAntonioCamurri,agreedtogivetheinvitedkeynotes. Thesetalkscovered di?erentfeaturesofmusicalexperienceanditsrelationtoArti?cialIntelligence, namelymusicalconcepts,composition,andperformance. Morespeci?cally,Jean- ClaudeRisset'stopicwasMusicalCompositionandArti?cialIntelligence:Some Precedents and Prospects,MiraBalaban'swas Structure and Interpretation of Music Concepts - Music from a Computational Perspective,andAntonio- murri'swasComputationalModelsofExpressiveGesture.Thepapersintheconferencedealtwithawiderangeofaspectsofcurrent research,includingstructural,harmonic,andreductionalmusicanalysis,pattern representationanddiscoveryinbothmonophonicandpolyphonicmusic,music perceptionofmelodyandrhythm,therelationbetweenmusicandnaturallan- age,similarityandcategorization,musicandintonation,musicalexpressionand performance,soundprocessing,soundclassi?cations,commercialapplications, andmusicontheweb. Acknowledgements. Theeditorsaregratefulforallthehelpofthosewho madethepublicationofthisvolumepossible,andespeciallyDarrellConklin, VI Organization PeterNelson,EmiliosCambouropoulos,AlisonPease,andMarkSteedman;also, JordanFleming,EwenMaclean,andRobertDow;andtheFacultyofMusicand theDivisionofInformatics,UniversityofEdinburghfor?nancialsupport. July2002 ChristinaAnagnostopoulou MiguelFerrand AlanSmaill OrganizingCommittee AlanSmaill(Chair) ChristinaAnagnostopoulou(Co-chair) MiguelFerrand PeterNelson AlisonPease MarkSteedman ProgramCommittee ?JensArnspang(Alborg) GillianHayes(Edinburgh) MiraBalaban(Ben-Gurion) HenkjanHoning(Nijmegen) MartaOlivettiBelardinelli(Rome) JukkaLouhivuori(Jyvaskyla) RensBod(Amsterdam) TodMachover(MIT-MediaLab) CarolaBoehm(Glasgow) GuerinoMazzola(ZurichandVienna) AmilcarCardoso(Coimbra) StephenMcAdams(IRCAM) EmiliosCambouropoulos(Thessaloniki)EduardoMiranda(SONYCSL-Paris) ElaineChew(USC) PeterNelson(Edinburgh) DarrellConklin(ZymoGenetics) FranccoisPachet(SONYCSL-Paris) IanCross(Cambridge) StephenTravisPope(SantaBarbara) RogerDannenberg(CMU) AlanSmaill(Edinburgh) PeterDesain(Nijmegen) MarkSteedman(Edinburgh) AnastasiaGeorgaki(Athens) DavidTemperley(Rochester) InvitedSpeakers MiraBalaban(Ben-GurionUniversity,Israel) AntonioCamurri(UniversityofGenoa,Italy) Jean-ClaudeRisset(LaboratoiredeM'echaniqueetd'Acoustique,France) TableofContents InvitedTalks StructureandInterpretationofMusicConcepts:Musicfroma ComputationalPerspective...1 MiraBalaban ExpressiveGesture...4 AntonioCamurri RegularContributions AGeneralParsingModelforMusicandLanguage ...5 RensBod TheSpiralArray:AnAlgorithmforDeterminingKeyBoundaries ...18 ElaineChew RepresentationandDiscoveryofVerticalPatternsinMusic ...32 DarrellConklin DiscoveringMusicalStructureinAudioRecordings...43 RogerB. Dannenberg,NingHu RealTimeTrackingandVisualisationofMusicalExpression ...58 SimonDixon,WernerGoebl,GerhardWidmer AutomaticClassi?cationofDrumSounds:AComparisonofFeature SelectionMethodsandClassi?cationTechniques...69 PerfectoHerrera,AlexandreYeterian,FabienGouyon SomeFormalProblemswithSchenkerianRepresentationsofTonal Structure ...81 TimHorton RespirationRe?ectingMusicalExpression:AnalysisofRespiration duringMusicalPerformancebyInductiveLogicProgramming ...94 SohIgarashi,TomonobuOzaki,KoichiFurukawa MimeticDevelopmentofIntonation ...107 EduardoReckMiranda InteractingwithaMusicalLearningSystem:TheContinuator ...119 Fran,coisPachet VIII TableofContents RecognitionofIsolatedMusicalPatternsUsingHiddenMarkovModels ...133 AggelosPikrakis,SergiosTheodoridis,DimitrisKamarotos AModelforthePerceptionofTonalMelodies...144 Dirk-JanPovel ControlLanguageforHarmonisationProcess ...155 SomnukPhon-Amnuaisuk EvaluatingMelodicSegmentation...168 ChristianSpevak,BelindaThom,KarinHothker CombiningGrammar-BasedandMemory-BasedModelsofPerceptionof TimeSignatureandPhase ...183 NetaSpiro ABayesianApproachtoKey-Finding ...195 DavidTemperley AuthorIndex...207 StructureandInterpretationofMusicConcepts: MusicfromaComputationalPerspective MiraBalaban DepartmentofInformationSystemsEngineering Ben-GurionUniversityoftheNegev,Israel. mira@cs. bgu. ac. il Researchinthe?eldofComputer-Musicisdirectedtowardsconstructionof computersystemsforcomposition,performanceofmusicaltasks,musictra- ing,signalprocessingandextensionoftraditionalmusicsounds,notationstudy, musicanalysis,storageandcommunicationofmusicaldata,andmusicinf- mationretrievalandclassi?cation. Computerapplicationsinmusicleadtothe developmentofcomputationalmodels. ExamplesaretheworksofWinograd, Ebcioglu,Laske,Cope,Bamberger,Berger&GangandOrlarey. Berger&Gang usearti?cialneuralnetworkstoinvestigatemusicperception,Orlareyetal. - troducethenotionofConcreteAbstraction(borrowedfromLambdaCalculus) asabasisfortheirpurelyfunctionalcompositionenvironment,whileWinograd usesaSystemicGrammarapproachborrowedfromLinguistics. Inspiteofrichactivityinmusicologyandincomputermusic,thereislittle research(ifatall)onthedevelopmentofacomputationalbasisforcomputer music(e. g. ,Schenker,Narmour,LerdhalandJackendo?).Mostcomputermusic systemsindeedhavehugemusicalknowledgeandintuitionsembeddedinthem, butcannotpointonunderlyingexplicitcomputationalbasis. Clearlythey- sumesomelevelofprimitives,andusedi?erentabstractionlevelsandkinds,but thesearenotanchoredincomputationalmodelsthataccountfortheirbasics, organization,usageandsemantics. Inasense,thissituationremindsoneofthe earlydaysofNaturalLanguageProcessing,whenalargeamountoflinguistic knowledgewasembeddedinhugesemanticnetworks. Inthistalkweproposetoinitiateastudyofcomputationalframeworks formusicknowledge. Theideaistomakethemusicknowledgethatunderlies computer-musicapplications(eitherintentionallyornot)explicit. Asastarting point,wesuggesttostudy music schemas. HereishowMusicologistDahlia Cohen(dept. ofMusicology,HebrewUniversity,Israel)describestherolethat musicschemasplayinmusicunderstanding: Listeningtomusic(perceptionandreaction)anditsmemorization,are cognitiveactivitiesthatdependon schemas thatarecreated,unc- sciously,atanearlyage. Theschemasrepresentrelationshiptypes- tweenmusicaleventsthataredictatedbyprinciplesoforganization,in variousabstractionlevels. Theschemastriggerexpectationsthatmay ormaynotberealized.Musicisdistinguishedfromotherartsinhaving thesystematicorganizationasthesolemeansfortriggeringvarioustypes ofreactions(whereasotherartsemploynon-artistic,non-organizational meansforreactiontriggering).