computationalmusicology a survey on methodologies...

Extrait de la Revue Informatique et Statistique dans les Sciences humaines XXIX, 1 à 4, 1993. C.I.P.L. - Université de Liège - Tous droits réservés.

Computational MusicologyA Survey on Methodologies and Applications

Lelio CAMILLERI

Résumé. Cet article traite de J'utilisation de l'i.nformatique dans les études de musique. Il présenteune introduction aux différents aspects de la musicologie computationnelle el aux récents développements dan!'> ce domaine. Il passe en revue les principales applications de l'informatique en musicologiehistorique ct en analyse quelques exemples et leurs implications. Ensuite sont décrites les applicationsde l'ordinateur aux éludes analytiques en cc qui concerne leurs aspects méthodologiques. Desexemples de systèmes d'analyse automatique sont expliqués avec les conclusions théoriques que l'onpeut cn tirer. L'article s'achève par une analyse rapide de développements cn\,jsageablc.~ dans lefutur.

KClwords: musicology, analysis, artificial intelligence, computational methodologics, musicaldatabases.

1. Introduction

Mots.clés: musicologie, analyse, intelligence artificielle, méthodologies informatiques, bases dedonnée.:; musicales.

Computer applications ta music caver a \Vide range of musical activities:sound processing, composition, psychoacoustic, music printing and musicology.For the applications concerning the lllusicologicai field a new heading wascreated: computational musicology. But what cOlnputational musicology doesmean? It is evident the analogy to computational Iinguistics but Ihe content ofcomputational musicology, in the field of music obviously, could be comparableboth of Iiterary and Iinguistic computillg. [n fact, computational musicologysludies concern not only with theoretical studies but also with the creation ofmusical repertoires in machine readable form, historieal and musical databasesand other. The computational musicology applications range from studies onmusical historical texts ta musico-cognitive studies; evidently, the degree of use

o Divisione Musicologica CNUCE-C.N.R.; Conservatorio di Musica L. Cherubini; Piazza delleBelle Arti, 2; 1-50122 Firenze (Italia).

Fax: + 39552396785 E-mail: [email protected]


52 Lelia CAMILLERI

and implication of the utilization of the computational methodologies variesdrastically from one application ta another.

The current situation of the computational musicology studies is weil established. Musicologieal studies using computer appear quite regularly in theinternational journals of the field. In addition, an important publication in whichthe major activities in this field are reported is published yearly by the Centerfor Assisted Research in the Humanities of Menlo Park, USA, since 1985, titledCompl/tillg ill Ml/sica/ogy. Another publication on this field is Com/mtet" ill Ml/sicResearcll, published by the Wisconsin Center for Music Technology, in whiehtheoretical and applicative issues are also viewed in respect ta their educationaluse.

A study group on musieal data and computer application has been established in the International Musieologieal Society and recently organized a roundtable on the New Methodologies in the Study of Melody at the IMS XlV International Symposium in Madrid in 1992. In addition, computational musicologystudies are weil represented in the two editions of the Conference on Computersin Music Research (Lancaster: 1988 and Belfast: 1991): in this conference themusical research part in the computer applications ta musie is emphasized ratherthan the compositional and sound processing ones.

ln the recent years various books (Marsden and Pople: 1991; Balaban et a/.:1992, Hewlett and Selfridge-Field: 1992) and journal special issues (Camurri:1990, Camilleri: 1992) in which aspects of computational musicology are treatedhave been published.

ln tllis artiele 1 will briefly sununarize the various fields of application andfocus on some examples in each topies. In same cases the examples rely uponworks carried out by myself and my colleague at my Institute.

2. Historicalmusicology

ln the field of historieal musieology we can single out two main areas:(1) IIlusieal repertoire in machine readable fOrIn, (2) databases of musical textand/or historical and bibliograpllical information, and thematie catalogues.

2.1. Musical reperloires in machine readable forlll

Il is self-evident, but often not sa obvious, the importance of the realizationof musical repertoire in machine readable fOrIn. The first implication of creating


CO.r-.1PUTATIONAL MUSICOLOGY 53

repertoires in machine readable form is the possibility to realize modern criticaledition at low cost. In fact, a musicologist can carry out ail the editing process byhimself, taking care of ail details and saving a considerable amount of moneyin the publication budget. By this way, several editions, of so called minorcomposers, have had the possibility to be printed because of the small budgetrequired. In this case the major importance is duc to the advent of musicalprinting programs. However, the important side etIect of using them is thefollowing one: if the printing program provide a readable and explicit encodingscheme then the amount of data representing the particular repertoire can beused, in principle, by other scholars who do not need to re-enter ail the musicinto another format. An experience on this application has been realized at theDivisione Musicologica CNUCE-CNR, Florence, where the critical edition ofthe Second Book of Madrigals by Luca Bati (Gargiulo: 1991) has bcen realizedby musicologist Piero Gargiulo. The files containing the 21 madrigals are storedin the computer facilities of our Institute at other users disposai. Ali the materialis encoded in Score code, the alphanumeric code of the printing program Score,which is plain ASCII, weil structured and with a defined and interpretablemeaning for each symbol used. Therefore, a musicologist who would Iike toanalyze these madrigals using computer can utilize the already encoded materiai.

Large musical repertoires in machine readable fonu have been created atthe Center for Assisted Research in the Humanities. One of these projects isaimed at encoding the complete works of J.S. Bach using an encoding schemewhich is alphanumeric but uses an internai representation which allows to storeail the facets, in addition to piteh and duration, of printed score.

A good example of the structure of several encoding schemes is found in the1987 DirectOlY ofCompl/ter Assisted in Ml/sicology (Hewlett and Selfridge-Field:1987) where the same musical incipit has been encoded by difl'erent systems andtheir facets are discussed.

2.2. Musical databases

Databases often represent the first approach of the musicologist who usesthe computer. Storing, retrieval and comparing itûonnation are operations thatmusicologists have done manually for years. The possibility to automatize atdifl'erent degrees these operations ean improve the search speed and the accuracyof the research.

As mentioned before, one can distinguish between two main kinds of musicrelated databases; databases of historical bibliographical or any other type of


54 Lelio CAMILLERI

textual content, and databases of musical information, containing musical text asthe primary source on whieh to execute searches and comparisons.

An interesting research dealing with the realization of a textual databasehas been carried out by Sandra Pinegar (1991) as a part of her doctoral thesisat Columbia University, USA. Thema is a textual database whose contentare transcriptions from thirly-seven thirteenth and early fourteenth centurymanuscripts of Latin treatises of mensurable musie. Thema main fenture isthat the transcription should accurately reprcsent the original lexl as in aphotographie facsimile, i.e. using the same line length, paragraph signs, andother elements. Thellla uses a system of numerieal tagging that encodes theabbreviations. !t allows a series of operations such as a comparison betweentwo more copies of the text, text columns, sections of the copy and between twoor more different texts. The materials of Thellla is transportable as ASCII filesin order to allOW different operations on them ranging from word processingto concordance indexing or word-crunching. Another similar example of texlualdatabase based on the same subject is the Thesaurus MusÎcal1un LatinaJllm, aproject realized by Thomas Mathiesen at Indiana University (Mathiesen: 1992).TML cncoded texts are not the original manuscripts but their modern editionsand the historical period treated is from Middle Age to Renaissance. TML hasan encoding scheme to store and represenl the musical incipits and the databasefacilities are available through the computer network. By this way, remote userscan easily access the database content and perfonn searches.

However, musicologieal text based databases are of different kinds ofcontent. They concern \Vith information on musical sources, newspaper musicalarticles in a weil defined period, or nota tiona1images of a partieular musical age,only to quote a few. Thus far, we have dealt with textual information. Anyway,musieal databases can have the musical texts as main part of their content andnot just the incipit as we have seen in TML.

Nonnally, these databases are analyticallY oriented, the musical text isstored with the purpose of performing some types of analytical searches on il.Therefore, the border between analytical programs based on comparative andquantitative techniques and the one based on databases is very fuzzy. Two veryimportant researches of this kind, based on very different repertoires, are relatedto the systems Scribe and Esac/Mappet.

Sctibe is a datalxlse of fourteenth ecntury music developed and maintainedat La Trobe University, Australia, realized by John Stinson (1992). !ts content isfonned by 4,408 fully encoded works. Scribe allows to enter, using an alphanumeric code, the chant melodies and displays them in mensurai and stail' notation.The database allows to execute search and operations with difl'erent criteria


COr-.-IPlYl'ATIONAL IvluSICOLOGY 55

and display the result in two fonns: (1) files whieh can be re-read with somecommercial database program, (2) printoutwhich are realized through a pIotterand realized in neumatic notation. The comparison of a corpus of Dominicanchant with 105 Cypriotic chants by means of Scribe facilities has yielded somehypotheses about the composition of late medieval chant (Stinson: 1992).

The work carried out at Essen University by Helmut Schaffrat and hisassocia tes produced a software package for the encoding, analysis, retrieval andcomparison of melodie repertoire of any kind. Il also comprehends a usefulprogram, Estaff, which allows fmther analytical operations and the display ofmelodies in staff notation.

The first point ta stress in respect ta the software of the Essen group is thatit is easy-to-use and tnily operationa!. The problem many software have is thatthey are operational only for the research group who conceive them, and thustheir use is limited. On the contrat)', this software package is used, in additionta the Essen group, by other scholars for different purposes.

The Essen software has been primaIY conceived for German folksongsand Chinese melodies but it couId be used ta process any kind of melodicrepertoires. Essen database contains more than 10,000 melodies. ln this briefdescription l would like ta bring out some features of this software. The melodyis represented in Esac notation, an alphanumeric code which uses numbers torepresent scale degree and lines ta represent duratian (each line added ta anumber indicates the multiplication of the basic duratian unit which is specifiedtogether with the tonie in the KEY field). In Figure l, a melody (Wal/derersNachtlied by Karl Collan) is displayed bath in staff notation and in the analyticaloutput realized by the program. The analytical output contains the whole melody,in the field MEL, together with otlter analytical information. As we can see,each analytical result and the melody itself are placed within square bracketsand specified by a tln'ee letter field. For instance, the field For (form of pitch)indicates the intelval sequence in each phrase labelling them witll a letter. Asequence denoted with 'av' indicates the first pln'ase labelled 'a' with a variation.Il is the same for the field FOR (Form Of Rhytlnn).

Another relevant field is the one labelled FOC (Form Of Contour) in whiehthe contour of each phrase is represented. A1phanumerie code is used ta indicatethe different kinds of contour and the comprehension and comparison of theseresults is facilitated by another program which allows their graphic display. Inthe staff notation example, realized by this program (EstaiT), the content of thefields Far, FOR and FOC are displayed on the right side. FOC field content isalso displayed in graphic notation.


56

lIm:lmrs !Iy..hlli~d

Lelia CAMILLERI

(0016 fûi fOR roc fin(

Fig.!.

COLLANUl/[13] 2S[10] 2M[14] 3S[3] 3M[0] 4P[1] 4A[0] 6P[1] 6S[3] 6H[3]

7S[0] 7M[0] 8P[0] 8L[0] SI[36] ABS[71] SA[24] LA[11]SB[23] 2s[14] 2m[28] 3s[6] 3m[1] 4p[3] 4a[0] 6p[0] 6s[0] 6m[0]78[0] 7m[O] Bp[O] 81[0] si(52] dt[-17] sd[42] Id[10]10[19] 20[18] 30[18] 40[17] LL[O] 00[18] ID[9] QD[O] TS[O] SD[B]

SL[7nLD[O] L3M[0] L3[0] L4[0] L4A(O] LG(O] L6M(O] L6[1] L7M(O] L7(l]M1(4] M2M(O] H2(7] M3M(O] M3(15] M4(8] M4A(O] H5(10] M6M(6] M6(21]

M7M[1] Hnl0] Ul(11] U211(0] U2(3] U3M(O] U3[1] HD(O] »GUT[Wanderers Nachtlied]REG (FinIand]SRC(Karl Collan (1828-1871) - Samlade Solosanger Vol.I

Edition Fazer]RHY(X_.XX_.XX_.X X _X_.X X__ .X_X_.X X O__J »REYle0016 16 0 3/4 ] 22[12]MEL[-6_.-71_.23_.4 3 _

3_.4 5 __ .6_6_.4 3 0 __6 __ .3_+1_.6 6_.6b6b __O__7 __ .6h_3_.2 3_.21 _6_.7hS_.64_3_3 __ 2 __6_.7 +1 __ .6_6b_6_7 0 __+1- __ 6 __ 6 __ 5 _+3 7 __ +2 __ +1 _+1 +1 __ +2 __ .+1_7_6_7 __ .6_5_.4 3 0 __+1- __ 7 _6_6 __ 5 _5 __ .43_.2 1 0 __ //] »

MOD[NINE-TONE-SKALEMAJOR?] RANGE(12] (-6,+3)CAD[336b1276+16361]ACC[-63_53_66_13_63_+17_+16_+3+2_+1+2_73_+16_1_]FOP (' a __b __ c __bs"d __ cW _e __ L_g__h __ L_hs_]FDR(Ja __b __bw_bw_c __ d __ e __ e __ f __bv_g __h __]FDC(JI __ ~ __ Sd -a ]

PHRAS (MAINLY \l1T1I DOWNBEAT]

CMT(Text by W. Goethe]

Another important feature of the Essen software is that ail the results arestored, as we have seen, in a database format and more precisely in the ASKSAM


COMPUTATIONAL MUSICOLOGY 57

one. Therefore, subsequent searches can be perfonned and results from differentrepertoires cao be merged and compared.

ln addition of these facilities, one of the programs of the package allowsthe display of some analytical results and of the melodies in staff notation; but italso selves as an analytical notepad where the analyst can single out interestingfragments, store them, retrieve and compare.

3, Analytical stndies

ln the lasttwo examples, and especially in the last one, we have seen as thedatabase application whose content is made up of musical texts are, de facto,analytically oriented. In fact, we can see the Essen group software useful fordiachronie comparison and for incipit storing also, but ils utilization is mainlyanalytical.

The use of computational methodologies to build up systems ta analyzemusic represents an important contribution for the music theOl'Y studies. Thiskind of application has twofold implications that, in certain cases, are bothpresent in the system: (1) creating systems to assist the musicologist in canyingout the analytical task, (2) studying the theOl'y and ils analytical system by ilsimplementation in a computer program.

As far as the first case is concerned, we can think of a particular kind oftheOlY, set-theory for example (Forte: 1973), whose implementation makes itsuse easier for the analyst (Castine: 1992) because of the series of operationswhose manuai execution is slow and can produce mistakes. Another exampleof such a kincl is the harmonie analysis program realized by Ferkova (1992). Illabels chords, individuates keys and regions and their relationships by means ofa functiona!theory of hannony implemented in the program.

The case in which the realization of a computational analytical system hasas a purpose the investigation of a theory also, can be weil represented by someworks using Artificial Intelligence techniques. 1 present some examples of thiscase later, but before 1 would like to describe two operational methodologieswhich play a different role in designing a computational system for musicalanalyticaltasks. Figure 2 displays the two types of methodology.

ln the right diagram, an homogeneous repertOlY is scrutinized and itsobselvation yields a set of rules wWch represents the structural features andregularilies of this corpus. The set of rules is then implemented in a computerprogram whose task is to generate some musies which presents the same stylistie


MUSIC

IMPLE~1ENTATION

RULES -+ THEORY

58

REPERTOIRE ~ OBSERVATION

/ t///

//

//

//

//

/1 _

Fig. 2.

r------111111111111111L _

Lelia CA1vffi...LERI

-+ THEORY

IMPLEMENTATION

ANALYSIS

ANALYTICAL RESULTS

or structural features of the corpus under investigation. In the case of wrongresults, i.e. a music with some structural features not stylistically pertinent withthe corpus, the corpus is re-obselved to individuate the missing features, the setof rules is re-fonnulated and the computer program is updated. The path can betravelled over again several times in order to have the proper results. In this case,the results are the rules which describe the characteristics of the repertoires; thecomputer is used only as a tool to assist the theorist in his research work. Theimplementation of the rules in a computer program selves as a mean to test thehypotheses formulated (rules) due to the fact that the computer do not performany Interpretation on them: thus, wrong rules give wrong results and it allows thetheorist to investigate the unexplained parts of his formulations. This approachis called generative due to the fact that the theory test process is realized bygenerating a piece of music.

In the other diagram, the left one, the content of the software is a theorywithits analytical system but it Selves to perfonn an anatysi, on a musical piece. Thus,the results are a series of analytical information on the piece, i.e. the feature ofits rhythmic structure, whose degree of pertinence demonstrates the value of thetheory in describing a particular structural feature. In the case of non-pertinentresults, the theOl'Y will be re-scrutinized and improved. Therefore, this kincl ofmethodology allows the building of applicative systems, performing an analyticaltask, but has also the 'side efl'ect' to corroborate or invalidate the theOlY whichit is based upon.

An example of generative grammar based system is due by the work ofBaroni and his associates (1984, 1992). The research work investigated melodicrepertoires such as melodies of Bach chorales, French chansons of XVII centmy,and Legrenzi's Cantatas. In the last repertoire the relationships among harmony,melody and text is also taken into account (Baroni et al.: 1992). The study ofthree difl'erent melodic repertoires has the purpose to bring out the relationships


CO~1PUTATIONAL MUSICOLOGY 59

between the theory of melody general facets and repertoircs peculiarities, andthe comparison of micro-structure elements.

This approach has been also applied to other repertories such as jazzimprovisation (Giomi and Ligabue: 1991), and on the lirst phase of Kippenand Bel study (1991) of tabla improvisations. Another type of knowledgebased musical system which could be placed halJ\vay between the two kindsof methodology described above is represented by the work of Ebcioglu onthe Bach's Chorales. Il is based on a large body of rules formalized after theobservation of a musical corpus. The system has the purpose to investigate andperforms two tasks: (1) the harmonization in Bach's Chorale style of a givensoprano voice, (2) the realization of the Schenkerian analysis graph in which thehierarchy among the melodic elements is represented.

Ebcioglu's Choral system (Ebcioglu: 1987, 1988) is also a good examplein using expert systems for musicological applications and possesses severalinteresting features. The lirst aspect is given by the application of multipleviewpoint strategy. Il consists in viewing, and thus harmonizing and analyzingthe chorale by means of dilferent approaches. Each viewpoint has its ownproduction rules, heuristics, and constraints. The live viewpoint are the chordskeleton, lill-in, time-slice, melodic string, and Schenkerian analysis view. Thelast part is the analysis part of the system and its main interesting feature liesin the integration of assumptions from Schenkerian, Lerdahl and Jackendofl',and Ebcioglu own theories on the analysis of voice leading and prolongation.Incidentally, Ebcioglu's work with the adoption of multiple viewpoint (whichunderlies dilIerent methodological approaches) is a very promising experienceon the possibility to integrate distinct theories strategies. The parallel he madebetwecn hcuristic and the personal unconscious musical knowledge, is a pointto deepen as to enlighten some interesting facets on the high-level musicalknowledge.

This system, beyond its interesting theoretical aspects has also an applicativeaspect; Choral harmonizes a soprano voice of a choral and produces also therelative Schenkerian analysis graph. In the following ligure are shown four barsof a Chorale. One is harmonized by Ebcioglu's program and the other one byBach. In addition, the Schenkerian graph with the hierarchical grouping of thesoprano voice is displayed.

Another work based on integration of theories and expert systems is the onefocused on musical grouping structure carried out by myself and my colleagues(Camilleri, Carreras, Duranti: 1990). In the study of musical segmentation, theIntegration of theories of partitioning (from micro level to macro level) cansolve some crucial problem in this subject. We have endeavoured to relate the


60

B c D

Lelin CAMILLERI

Fig. 3.

A = intcrvalleap; B :=: parallelism; C = parallclism; D = dynamics.In the case A the interval leap indicatcs a group boundary. In the case B the t'Vosegments (l'rom 84 to 04 and l'rom A4 10 F4) have a sunilar intervallic structure andthe rule of parallelism is applied to define the groups. ln the cases C and D theprogram chooses the solution by weighting the cule priorit)'. Numbcrs and applies inthe fifst and second level fuie 5 (dynamics) and rule 8a (paraUelism), respectively.

rules extracted and reformulated from Lerdahl & JackendolI's theory (1983), themore exhaustive framework in this subject, to rules derived from psychologiealinvestigation, musieal semiotic analysis (especially taxonomie one) and other.

This research work starts from the aforementioned considerations andpursues two purposes: (l) the creation of an analysis environment, to beimproved and amplified as to allow other kinds of analysis (tonal harmonieanalysis), (2) the testing and improvement of the content of the theOlY ah'eadydeveloped, and the verification of the possibility of an Integration of diIrerenttheoretical assertions.

As 1already mentioned, from a general theoretical point ofview, concerningmusical segmentation, a rule system has been fonnulated starting from Lerdahland JackendoIr's theory (for the metric and rhythmic aspects), and integratingit with other assertions belonging to diIrerent approaches. This perspective is,on our view, very important bath for the modelisation and implementation ofmusical kllowledge. If we think of musie analysis not only as the labelling andclassification of musieal entities but as a process of reasoning about musie (thusplaced at a higher cognitive level) we should take into account the way a piece canbe scrutinized by means of several analytical concepts integrated iuto a generaiframework. In addition several disciplines developed in the latest twenty years,for instance scnl.Îotic of music, music perception and cognition studies, new musictheories share more fundamentai notions than is nonnally assumed, such as oneof symbol.

The rule system formula tes a measurable definition of group boundariesby weighting the role of several musical parameters (interval, metrical position,dynamies, and so on). This metrie should represent the perceptual space in whiehthe parameters are mapped out. Studies on perception of the grouping structurewere employed in order to include facets detennining group boundaries like


COl\1PUTATIONAL MUSICOLOGY 61

pivot-note, melody direction, and ot11ers. Musical semiotics hypothesis on "iconicsimilarities" and recognition of basic musical units were used ta enhance the notion ofparallelism in the attempt of taking into account the indetenninateness ofsegment likeness. Furthermore the notion of parallelism, completely undefinedin Lerdahl & Iackendoll"s theory, is formulated by c1assifying through rules somesimilarities related ta musical parameters (interval contour, rhytillnic figure, andsa on). Within theseeategories, the system decides whether two musical segmentsinclude some mutual relationships and, thus, belong ta a group of higher levelor a segment is a transformed repetition of the other.

From the analytical point of view, the program output presents the hierarehical structure of the group subdivision into di/t'erent levels together with theelements (sequences of notes and sequences of groups) contained in each oneof them.

Sdlènkèl;arl f,I~p/Jof Chülal1l0. 128 lea1iH<Î by ltl~ CHORAL plOg[;l.Jll

Fig. 4.

Figure 4 shows an example of a grouping structure of Schubert's Lied DasWalldem as detected by the system. If we carefully examine the segmentationat certain points we can discover some useful indications on the behaviour ofthe rule system. The point marked with A shows that the system infers thatthe interval leap represent a group boundary between two groups at lowerhierarchicallevel. Point Band C represent the decision of the system due ta theevident parallelism of the two musical segment at a higher level. At last, point 0shows an activation of the rule which takes into account a change of dynamics asa relevant sign of a group boundary.

11le ta01 selected for the implementation of the expert system was the shellESE of IBM. The shell accepts production rules of the type IF/THEN, and theforward and backward chaining inference strategies; these two strategies aredynamically activated during the processing of the knowledge base by recallingspecial control statements. Various types of control statements arc available andprovide the means for accessing externa1 data, for cOllullunicating \Vith the userthrollgh an interface that may use graphics, for directing and conditioning theexecution process.

An extension of this work is the realization of an expert system prototypefor the harmonic analysis using another part of the Lerdahl & Iackendoll' theOlY,the time-span reduction (Camilleri, Carreras and Giomi: 1992).


62 Lelin CAMILLERI

An analytieal work based on an ethnie repertoire, tabla improvisation, presenting interesting features is carried out by Kippen and Bel (1989). The groundfacet, due to the peculiarity of the repertoire, is the attention in avoiding theinclusion of the personal bias of the analyst, from his western based perspective,to the assertions of the native informants. For this reason the system is suppliedby music examples of improvisations perfonned by the infonnants and from theseit abstracts the regnlar repertoire featmes by means of an inference engine. Thetesting phase is realized through improvisations realized by means of the rulesabstracted by the system from the examples.

4. Further perspectives and conclusions

ln this brief sUlvey, we have seen the various applications of computer tomusieology and described their methodological implications, mainly in respecttoanalytieal studies. ln this last field the applications of the latest computationalparadigm, i.e. neural network or connectionist systems, can provide a relevanttool to by-pass a crucial problem: the representation of the musical score insymbolic format. ln fact, the musicaltext supplied to a large number of analyticalprograms is not the real representation of what is heard by the analyst but it isthe codification of its sign in the score. In this case, the sounding dimension, andail ils relationships wilh the musieal structural features, is los!. System based onconnectionist architecture ean provide a useful framcwork to 'interface' soundil1gproperties with high level ones, i.e. harmony. A very promising research on thissubject is carried out by Leman (1992) who tries to discover harmonie featuresof a piece of musie starting from its sounding content. The input of the systemis taken from CD and the data arc sent to an 'car moclel' which computes therelevant information subsequently sent to a system whieh related them to theharmonie structure. By this way il is possible to study the relevance of therelationships between psychoacoustical properties and their implications in theharmonie structure of a piece.

In the other field of computational musieology, an important step to improvethe data acquisition is given by the realization ofsystem fol' the Optical CharacterRecognition of musie. Several research works have been developed on thissubject but none of them has reached the level of applicative use (Hewlett andSelfridge-Field forthcoming presents an ex11austive survey on OCR for music).Another aspect of this problem is the code with whieh the scanned musie will bestored.


COMPUTATIONAL j\,fUSICOLOGY 63

To solve this and other problems of utilization of music corpora in machinereadable fonn the creation of transcoding program is desirable: il could allow theresearcher to casily access the corpora wrilten in the major encoding language.

We have also seen as the musicological work, in aH its aspects, can he carrjedout by using cOlnputers. Musical databases can help musicologists either to makecross references ta a corpus of documents and ta scrutinÎzed similar features ondifferent musical corpora. Moreover, music printing programs allow musicologistto execute ail the editing task by themselves and thus make the realization ofcritical editian easier.

In the field of analytical studies, besides its applicative use, the computerapplication can bring about an investigation of the theory used. Therefore, inthis subject the computer is not only a tool but a methodological instrumentwhose use yields effects on the improvement and re-definition of the theoreticalformulations. Computational musicology is actually a field in continuous develapmentwhose effects, În tenns ofeasy-access ta musical cm·para, analytical taaIs,database realization and other, wjll represent an important contribution ta themusical studies for the years to come.

BibliographyBALABAN (M.), EBCIOGLU (K.), LASKE (O.): 1992, eds. Ulldmlalldillg Mllsic ",ilh A.I.

(Cambridge Ma.: The MIT Press).

BARONI (M.) et a/.: 1984, nA Granunar for Mclody. Relationslùps between Melody andHarmony", in BARONI (M.) and CALLEGARI (L.) cds., Musica/ Grammars andComputer Ana/ysis (Olschki).

BARONI (M.), DALMONTE (R.), JACOBONI (C): 1992, '1\nalysis via Synthesis: theProgram Legre" in DALMONTE (R.) and BARONI (M.) eds., Proceedings ofthe 2ndEuropean Conference ail Music Alla/ysis (11:ento: Università di n'cnto).

CAMILLERt (L.), CARRERAS (E), DURANT! (C.): 1990, '~n Expert System Prototypefor the Study of Musical Segmentation", Illterface 19 (2).

CAMILLERI (L.): 1990, "Computational Theories of Music: Theoretical and ApplicativeIssues", in MARSDEN (A.) and POPLE (A.) cds., Computer AIodels and Represelltafions in Music (London: Academie Prcss).

CAMILLERI (L.), GIOMI (E), CARRERAS (E): 1992, "Time Slices for Time SpanReductions: An A.I. Environment for the Analysis of Tonal Harmonie Structure", inHEWLETT (\Y.) and SELFRtDGE-FIELD (E.) eds., Complllilig ill Mlisicology (MenloPark CA: CCARH).

CAMILLERI (L.): 1992, cd. Mitlds alld Machilles 2 (4) [Special issue on Music andCognition].


64 Lelio CAMILLERI

CAMURRI (A.): 1990, ed. IlIlelfacc 19 (2-3) [Special Issue on Artificiallntelligencc andMusic].

CASTINE (P.): 1992, "Contemporar)' Music Anal)'sis Package for Apple Macintosh(CMAP)", in DALMONTE (R.) and BARONI (M.) cds., Procee<lillgs of Ihe 211<1European Conference 011 Alusie Analysis (TI'cnta: Univcrsità di 11ento).

EBCIOGLU (K.): 1987, "Report ou the CHORAL Project: An Expert System for Harmonizing Four-Part Chorales", IB1\1 /Illemal Report (TJ. \Vatsoll Research Center,Yorktowu NY).

EBCIOGLU (K.): 1988, ''An Expert System for Chorale Harmonization", Computer MusicJOllm1l1 12 (3).

FERKOVA (E.): 1992, "Cach: Computer Analysis of Classical Hannony", in DALMONTE (R.) and BARON! (M.) cds., Proceedings of the 2nd European Conference011 Music Alla/ysis (11ento: Università di 11:cnto).

FORTE (A): 1973, The ShUClUre of Atonal Music (New Haven: Yale University Press).

GARGIULO (Piero): 1991, Luca Bati Mml1iga/isla FioreJllino (coJlI'Edizione Modema deiSecolldo Libro de Madligali a cinque voci) (Florence: Olschki).

GIOMt (E), LIGABUE (M.): 1991, "Computational Generation and Study of Jazz Music",IlItelface 20 (1).

HEWLETT (W.B.), SELFRIDGE-FtELD (E.): 1987, eds. DireclOl)' of Compllter AssistedResea,.c" ill Mlls/cology (Menlo Park CA: CCARH).

HEWLETT(W.B), SELFRIDGE-FIELD (E.): 1991, "Computingin Musicology 1966-91",Compltlers and Ihe Hllmanities 25.

HEWLErr, (W.B.), SELFRIDGE-FIELD (E.): 1992, cds. Complltil1g il1 Musicology (MenloPark CA: CCARH).

HE\VLETI' (WB.), SELFRIDGE·FfELD (E.): forthcoming, cds. Compuling in Nfllsicology(Menlo Park CA: CCARH).

KIPPEN (J.), BEL (B.): 1989, "The Identification and Modelling of a Percussion 'Language' and the Emergence of Musical Concepts in a Machine-Iearning ExperimentalSet-up", Complllers and Ihe Hllmanilies 23.

LEMAN (M.): 1992, "The Theory of Tone Semantics: Concept, Foundation, and Application", Minds and Machine 2 (4).

LERDAHL (E), JACKENDOFF (R.): 1983, A Generalive Them)' of A111Sic (CambridgeMass.: MIT Press).

"MARSDEN (A.), POPLE (A.): 1991, Complller "'fodels and RepresenlalioJls ill lV/usic(London: Academie Press).

MATHIESEN ('I:): 1992, "The Thesaurus Musicarum Latinarum", in HEWLETT (w.) andSELFRIDGE-FIELD (E.) eds., Complltillg ill Musicology (Menlo Park CA: CCARH).


CO~1PUTATIONAL MUSICOLOGY 65

PINEGAR (S.): 1991, Textllal mul COllceptual Relationships Among Theoretical JYdtings 011

Mellsumble Music dwillg the Thù1heenth and Early Fow1eenth CClltlllies, PhD. Dissertation, Columbia University.

SCHAFFRATH (H.), "The Retrieval of Monophonie Melodies and their Variants:Concepts and Strategies for Computcr-Aided Analysis", in MARSDEN (A.) andPOPLE (A.) cds., ComplIters Representations alld Models ill A111Sic (London: Academie Press).

STINSON (J.): 1992, "The Scribe Da/abase", in HEWLETT (W,) and SELFRIDGEFIELD (E.) eds., COlI/pl/tillg ill MI/sic%gy (Menlo Park CA: CCARH).

computationalmusicology a survey on methodologies...

Documents