H        O        M        E

Chapter 2: USING THE PHONEMIC STRUCTURE OF A POEM AS THE BASIS FOR MUSICAL COMPOSITION.

The note and phoneme relationship

Using the rules of a language in poetry

Similarities and differences between phonemes and notes

Two approaches to phoneme and note relationships, Nattiez and Wishart

Creating a musical composition from the phonemic structure of a poem

Process 1: Translation of the text to phoneme symbols

Process 2: Phoneme symbol to number conversion

Process 3: Table lookup of note data using an ID number

Process 4: Creating the source cantus firmus

Process 5: Cantus firmus to core melody modification

Process 6: Composing a polyphonic texture from the modified melody

Process 7: Playback

EVALUATION

Chapter 2

USING THE PHONEMIC STRUCTURE OF A POEM AS THE BASIS FOR MUSICAL COMPOSITION

                This chapter looks at the process taken in deriving musical compositions based on the phonemic string of the poem ZOOMING IN, by Alex Skovron. The phonemic string is translated into MIDI information which is manipulated through algorithms built in Opcode's MAX environment. The processes shown here can be used with any text; ZOOMING IN was chosen for illustrative purposes only.

ZOOMING IN                                        ©Alex Skovron 1991

dot \ sphere \ planet \ continent \

coastline \ shadow \ city \ chess \

block \ building \ window \ chamber \

doorway \ table \ figure \ face \

eye \ iris \ pupil \ blood \

cell \ atom \ darkness \ dot

 

face \ forehead \ bone \ blood \

matter \ motion \ energy \ bliss \

    energy / hatred \ terror \ darkness \

        energy / light \ bliss \ dot

Skovron describes ZOOMING IN as a sonatina. It is part of a body of works using the same structural framework.

The note and phoneme relationship

                The focus here is in using the surface structure of a textual expression, in this case a poem, as the basis for musical composition. That is, applying the rules governing the relationships of the phonemes used in a poem to create the background structure from which a set of musical compositions are generated. To this end I look at the sonic aspects of music and language that can be represented with a notation, which is in this case phonemes and notes.

                The notation of phones is highly developed as a method for describing speech sounds. A subset of phones called phonemes defines the more significant aspects of the phonetic pool of a given language. This subset is used as a basis for analysing the relationships between speech sounds in a given language, discovering the rules that underlie the language and then, if wanted, generating new phoneme groups that fit within the given language.

                The notation of music is also highly developed as a method for describing the sounds to be produced by a musical instrument. Because musical notation is not instrument specific, (that is, the same note can be played on a variety of musical instruments) it does not give much information about the actual sound to be produced. By contrast, phonetic notation is instrument specific, in that it refers only to the voice. Musical notation can, however, be used as a tool in analysing the relationships of pitches within musical compositions. It is in the area of analysis that the two notation methods share common ground.

Using the rules of a language in poetry

                The sound of the poem can be quite musical. This is obvious when the listener either disregards the meaning of the text or listens to a poem in a unknown language. The sonic structure of the text is an integral part of what the poet is expressing to the listener and the poet's specialised attention to this is essential, I believe, in order for the text to fit within the realm of poetry. However, the poet's attention to the sonic structure is often obscured by the listener’s focus on the lexical meaning of the text. This attention is seen in the poet's use of alliterations, rhymes and rhythms to enhance the intended meaning of the text. It is these three aspects which give the musical element to poetry.

                Each language has a finite set of rules regarding the relations of the sounds within it and these rules must be adhered to in order for the language to communicate effectively; therefore the poet is forced to use alliterations, rhymes and rhythms within the language being used. This means that the sonic structure of the poem, an important expressive device, is bound by the lexical meaning of the text. It can be said, when considering only structural aspects, that a composer of music has an easier job in that he or she does not necessarily have to contend with reconciling two methods of communication, the lexical and the sonic.

                These boundaries do not affect areas of textual work such as concrete poetry, where the emphasis is on the appearance of the text in enhancing the lexical meaning, or sound poetry, where the emphasis is on the sounds of the text rather than the lexical meaning.

Similarities and differences between phonemes and notes

                The uses of notes and phonemes are, for the most part, quite different. Phonemes are a set of universally accepted and understood symbols used to describe the sounds of a language as it is spoken. These symbols are used to describe a language for analysis. In the act of musical composition the note is often used as a starting point of the composition from which the instrumentalist generates the music. This process can be seen in the composition of serial music.

                Most instrumental music can be described as an ordered succession of discrete sounds. To the casual listener, hearing music as a coherent set of musical phrases, this may not be immediately obvious. However these phrases are made up of discrete sounds, each of which can be observed through the use of a visual representation, such as a score; or, in the case of electronically realised music, a frequency or amplitude against time graph. Both these forms of representation, score and graph, come notionally after the fact.

                Representing something in a medium other than its own creates an abstract and symbolic representation. In this case a purely aural form of expression is represented in a purely visual way. This creates a continuum from an abstract, conceptual point, where the object exists only as a description, to an actual point, where the object exists as a finite, describable, entity.

                Notated music has an interesting place in the abstract to actual continuum. The set of symbols used to describe a composition can also be used to generate it. This can be seen in serial composition, where the tone row and its subsequent music may grow from an abstract, non-musical concept, such as translating the birth dates of the composer's children into interval sets. Another approach could be where the composer observes interesting visual relationships between notes written on a stave, which he or she then develops into a musical composition. In these two cases the concept of the composition is developed from and dependent on the symbols used to represent it. On the other hand it is possible for the composer to notate a melody that springs whimsically from their imagination and then develops into a serial composition.

Two approaches to phoneme and note relationships, Nattiez and Wishart

                Jean-Jacques Nattiez describes Pierre Schaeffer's "morphology of sound objects É [as being] founded on a descriptive inventory of their characteristics"[1]. This inventory is similar to the inventory of descriptions used in phonetics, in that the inventory comes after the fact. Nattiez discusses this need for a well defined set of graphic symbols, similar to the symbols used in phonetics, as being essential for the musical analysis of non-scored music. For anyone attempting to analyse electronic/computer music a set of universally understood descriptive symbols would be useful.

                Later Nattiez describes the phoneme and note as "discretized" units rather than discrete units[2]. This is true in that each phoneme is removed from its context and has very little meaning on its own, just as a note has very little meaning when removed from the context of a piece of music.

                Trevor Wishart draws relationships between the characteristics of music and language, saying that:

the melodic stream is pitch-disjunct and may be articulated by timbrel colouration. [And that the] language stream is timbre-disjunct and may be articulated by pitch inflections[3].

He goes on to compare the timbre fields of languages with the harmonic fields of music, equating the chromatic field of tonal, tempered instruments with the phoneme field of a single language and the harmonic field of non-tempered instruments with the phoneme field of other languages[4].

                Wishart offers a brief analysis of an extract of Kurt Schwitter's Ursonata[5], which can be described as a sound poem or, by looking at the structure of the text in a musical sense, as a monophonic sonata. In his analysis he looks at the shifting use of vowel and consonantal "resonances". After the analysis Wishart goes on to say:

All this differs from our perception of field characteristics in pitch-lattice music (apart from the obvious pitch stream/timbre stream distinction) in a number of ways which are, however, not intrinsic to text sound composition. First of all there is no counterpoint or chorusing. Secondly, there is no indication of rhythm (which might however be implied from the printed spacing) or tempo. Adding these, and other, dimensions we can imagine a sophisticated contrapuntal art based on the articulation of a multi-levelled timbre or timbre-motif (possibly phonemic) field structure[6].

                Nattiez and Wishart's approaches deal with the sonic relationships between the phoneme, as the "discretized" unit of language, and the note, as the "discretized" unit of music. In doing this they create a continuum of articulated sounds or utterances in which language, as it is heard, and music, as it is heard, exist. This continuum is shown in Figure 2.1.

Figure 2.1 Continuum of all sounds

                The relationships drawn by Nattiez and Wishart are to do with the audible surfaces of music and language. This is a very tangible relationship and treats the gamut of articulated sounds as a possible pallet for musical or textual based sonic composition. While increasing the possible sound pallet neither Nattiez nor Wishart move into the broader area of the possible composition tools made available by the note/phoneme relationship; one of these tools is discussed below.

Creating a musical composition from the phonemic structure of a poem

                The idea that music and language have useful commonalities at the structural level has been of interest to composers and analysts over the history of music. One of the early uses of language structure as an aid in the composition of music was applied by Guido of Arezzo[7] in the early tenth century. Here Guido used a part of the phonemic structure of a text as a structural device when composing plainsong. He supposed that if the text itself was well constructed it followed that music built on that text would also be well constructed. Guido’s system was to attach a number of pitches to a vowel and then to select from the pitches available to that vowel when it is encountered in the text. This table is shown in Table 2.1.

Table 2.1. Guido's table for matching vowels to pitches ^[1]:

vowels:	a	e	i	o	u
pitches:	G	A	B	C	D
	E	F	G	A	B
	C	D	E	F	G
	A

                While Guido's process does not account for rhythm, amplitude or timbre it is a very practical approach for the writing of plainsong. The style of plainsong composition is rooted in the voice and defines the vocal idiom in Western music, therefore attending only to the vowel sounds of the text is appropriate as these are the sounds most useable in singing.

                The approach I am taking here expands on Guido’s method by including the aspects of rhythm, amplitude and timbre. This is done by applying a unique MIDI[9] note-event to each unique phoneme used in the poem ZOOMING IN. For example: when the phoneme /d/ is encountered in the poem it triggers the MIDI note number 23 with a MIDI velocity of 12, a duration of 28 units of time and an inter-onset time of 41 units of time. A unit of time can be anything from a millisecond to however long the composer sees as appropriate for his or her composition.

                 These note-events are created using the computer algorithm "ROWMAKER". How these note-event attributes are generated is explained under Process 3 below. The duration and inter-onset times can be multiplied to fit a range more acceptable to the compositional requirements of the piece.

                By using this process the phoneme string of the poem forms a single line melody, or cantus firmus, which is used as the structural core of the compositions. This cantus firmus is filtered through, and enhanced by, two computer algorithms, called "COMPOSE" and "CANON", to produce the final compositions.

                Table 2.2 below shows the processes that are used in the left column, and the reasons and the methods for using these processes in the right column. Each section of the table is explained in greater detail further on in the chapter.

Table 2.2 Processes and methods used in the COMPOSE and CANON algorithms.

PROCESSES	METHODS
Process 1: Translation of the text to phoneme symbols.	A set of symbols using keyboard characters available in any computer is used here.
Process 2: Phoneme symbol to number conversion.	Each phoneme symbol is assigned a unique ID number. The benefit of using ID numbers is that they are more suitable for manipulation by computers. They can be assigned to any type of event, such as a graphic event or a sample playback, not just a note-event.
Process 3: Table lookup of note data using ID numbers.	The ID number is used to select a unique pitch, velocity, duration and inter-onset time from a table, thus creating a unique note-event for each phoneme character. The pitches, velocities, durations and inter-onset times are created using an algorithm called ROWMAKER.
Process 4: Creating the source cantus firmus.	The list of phoneme characters is stepped through in the order in which they appear in the poem, triggering their accompanying labels which in turn trigger each note-event. This succession of note-events produces the cantus firmus. It is this core melody which is used as the structural basis for the compositions.
Process 5: Cantus firmus to core melody modification.	The cantus firmus is fed into the COMPOSE algorithm. Here the composer applies constraints to the pitch, velocity, duration and inter-onset time to produce a modified melody.
Process 6: Composing a polyphonic texture from the modified melody.	The modified melody is fed into the CANON algorithm, which creates four other related voices. Adding a counterpoint creates a harmonic and rhythmic context for the melody.
Process 7: Playback.	The melody and the other related voices are sent to a MIDI instrument for playback.

Process 1: Translation of the text to phoneme symbols

                The translation of text to phoneme symbols used here makes the translation of speech sounds into a set of symbols that are easily used in a computer MIDI environment as simple as possible. I use the set of phones given by Alfred Blatter[10] as the set of speech sounds. The symbols I chose use the standard International Phonetic Alphabet (IPA), as given in Blatter, as much as is practical. Where I deviate from the IPA I have tried to use symbols which relate in some way to the speech sounds heard.

                The reasons for choosing this set of characters instead of a set of standard phonetic characters, such as the IPA are that:

                (1) no additional software is required;

                (2) a minimum number of keystrokes is required for each character. For example: the key strokes required for the symbol /ä/, representing the /ir/ sound in bird, are option-shift–R; and,

                (3): the ASCII[11] numbers for each keyboard character can be easily translated into MIDI/computer information. For example: the ASCII number for /ä/ is 228.

                Many decisions must be made when transcribing a written text into symbols that represent the sounds of the text. The transcription used here reflects the way I speak the English language (with an Australian accent), and the way the poem forces me to speak that language. The poet's use of alliteration, and the positioning of the text on the page suggests to me a steady, rhythmical reading style. My choice of phonemes is based on this style.

                The alliterations, and positioning style, can be seen in the first and second lines:

                                dot \ sphere \ planet \ continent \

The words dot, planet and continent are similar sounding, they each begin and end in stop plosives[12], the vowels and the nasals are mostly produced in the front of the mouth. Sphere has none of these attributes, it is made up of continuant fricatives and the vowels are produced in the back of the mouth.

                The second line:

                                coastline \ shadow \ city \ chess \

uses only three plosives, shifting the predominant consonantal sound from stop plosives to fricatives and the predominant movement in the mouth transfers from front to back.

                A translation of the poem into the phoneme symbols used here is shown on the following page and a full list of each phoneme, its sound and its symbol is given further below in Table 2.3.

 

dot \ sphere \ planet \ continent \

dot---sfEr---planet---kontinent---

coastline \ shadow \ city \ chess \

kOstlIn---SadO---siti---Ces---

block \ building \ window \ chamber \

blok---bildiN---windO---CAmbP---

doorway \ table \ figure \ face \

dHwA---tAbl---figP---fAs---

eye \ iris \ pupil \ blood \

I---Iris---pUpil---blud---

cell \ atom \ darkness \ dot

sel---atom---dRknes---dot---

 

face \ forehead \ bone \ blood \

fAs---fHhed---bOn---blud---

matter \ motion \ energy \ bliss \

matP---mOSn---enPji---blis---

    energy / hatred \ terror \ darkness \

      enPji---hAtred---terP---dRknes---

          energy / light \ bliss \ dot

          enPji---lIt---blis---dot---

The three dashes between each translated word refer to the whitespace backslash whitespace used in the poem's layout.

Process 2: Phoneme symbol to number conversion

                This conversion is done by simply selecting an upper or lower case alphabet character for each phoneme and using the ASCII number of that character minus 53. The benefits of subtracting 53 from the ASCII number are:

                (1) it puts the labels in the lower end of the MIDI spectrum, allowing the labels to be used as MIDI information in their own right.

                (2) the lower number makes the labels easier to manipulate when ordering them for playback. For example: the symbol A is attached to the label 12, which means it can be used as MIDI note C.

                The symbol, ASCII number and identification number attached to each phoneme are shown in Table 2.3 below.

Table 2.3 Symbol, ASCII number and ID number attached to each phoneme

Phoneme sound FRONT VOWELS	Symbol		ASCII#		ID#	Phoneme sound NASALS	Symbol			ASCII#	ID#
ee -seed	E*			69	16	m - mow	m			109	56
i - slid	i			105	52	n - no	n			110	57
a - spade	A*			65	12	ng - sing	N*			78	25
e - sled	e			101	48	DIPHTHONGS
a - had	a*			97	44	o - no	x*			120	67
a - lamb	L*			76	23	ou - pound	W*			87	34
STOP PLOSIVES						ai - pail	B*			66	13
t - to	t			116	63	i - pile	I*			73	20
p - pat	p			112	59	oy - toy	Y*			89	36
d - do	d			100	47	SEMI VOWELS
b - bat	b			45	45	w - witch	w			119	66
g - gone	g			103	50	wh - which	M*			77	24
c - cast	k			107	54	y - you	y			121	68
BACK VOWELS						l - law	L			108	55
a - palm	R*			82	29	r - raw	r			114	61
o - hot	o*			111	58	CONTINUANT FRICATIVES
aw - paw	H*			72	19	f - file	f			102	49
oo - look	K*			75	22	v - five	v			118	65
oo - boot	U*			85	32	th - thy	T*			84	31
o - float	O*			79	26	th - bath	F*			70	17
						s - sue	s			115	62
CENTRAL VOWELS						h - hat	h			104	51
ir - bird	D*			68	15	ss - mission	Z*			90	37
er - brother	P*			80	27	s - vision	J*			74	21
u - mud	u*			117	64	z - zip	z			122	69

* These symbols are not used by the IPA.

The speech sound /Q/ is used in the process shown here, not the traditional two phoneme symbols /k/ and /w/.

Process 3: Table lookup of note data using an ID number

The note-event table was created using a MAX[13] patch called "ROWMAKER". "ROWMAKER" generates a random set of n unique note-events.

Each note-event consists of four aspects:

                (1) a unique pitch, as opposed to pitch class, for example, the pitch  C2 is distinct from the pitch C3;

                (2) a unique velocity, the velocity of a MIDI keyboard key when struck (MIDI velocities 0 to 127 are used) ;

                (3) a unique duration, the time length of a note; and

                (4) a unique inter-onset time, the time length between note attacks.

                In Table 2.4 below many ID numbers and their note-events are not attached to phoneme symbols. These ID numbers and note-events are not used in the pieces presented here. The negative numbers -21, - 200 and -200 are applied to the ID number, pitch and velocity used for white space and punctuation to ensure that the equivalent of a rest is generated when punctuation or white space is encountered in the text.

Table 2.4 ID numbers and their note-events

Phoneme symbol	ID number	Pitch	Velocity	Duration units	Inter-onset time units
A	12	22	1	66	53
B	13	3	41	17	52
C	14	64	30	4	22
	15	17	39	8	40
E	16	61	33	16	34
F	17	11	26	25	29
	18	54	6	48	36
H	19	20	16	23	17
I	20	45	2	62	21
J	21	16	65	47	27
	22	56	47	54	58
L	23	52	9	19	11
M	24	15	56	20	9
N	25	69	13	50	39
O	26	27	62	52	30
P	27	31	45	69	32
	28	66	10	33	55
R	29	26	52	55	42
S	30	49	49	6	33
T	31	44	40	38	6
U	32	47	19	21	13
	33	21	29	39	20
W	34	29	42	34	31
	35	25	38	26	2
Y	36	59	58	61	43
Z	37	48	11	41	62
	38	13	57	7	18
	39	12	18	58	24
	40	36	35	35	26
	41	38	46	24	51
	42	58	63	56	4
	43	1	14	2	37
a	44	43	43	10	38
b	45	51	55	15	68
c	46	5	25	51	0
d	47	23	12	28	41
e	48	67	0	3	67
f	49	63	5	13	44
g	50	6	7	45	60
h	51	35	61	1	1
i	52	41	69	42	61
j	53	46	48	0	3
k	54	8	66	14	49
l	55	50	17	44	23
m	56	30	8	59	16
n	57	0	36	60	28
o	58	18	23	11	12
p	59	9	44	31	5
	60	4	50	46	50
r	61	62	59	64	59
s	62	10	68	9	48
t	63	2	31	32	63
u	64	65	34	57	56
v	65	33	53	37	65
w	66	7	64	18	7
x	67	40	3	63	15
y	68	32	67	30	46
z	69	55	20	22	69
Punctuation and white space	-21	-200	-200	10	10

Process 4: Creating the source cantus firmus

                To create the source cantus firmus a phoneme ID number is attached to each note-event in the table of note-events shown in Table 2.4.

                This composition process borrows heavily from middle to late twentieth century serial processes in that aspects of each note are pre-ordained and immutable. However, the pieces do not follow the serial method of cycling through a series of note-events. Instead, the note-events which make up the source single line melody are selected from the pool of note-events shown in Table 2.4. As each phoneme ID is encountered the note-event attached to that ID number is triggered. In this way a single line melody, referred to here as the cantus firmus, is created.

                Table 2.5 shows the order in which each phoneme symbol, and therefore each ID number and note-event is triggered. This table lists the first two lines of the poem; the entire list can be found in Appendix 2. A space of three rests falls between each word of the poem regardless of line breaks. Lines 175 to 180, shown in Appendix 2, refer to the break between stanzas and this is represented by six rests.

Table 2.5 ID numbers attached to note events.

Event order	Phoneme symbol	ID number	Pitch	Velocity	Duration	Inter-onset time
1	d (dot)	47	23	12	28	41
2	o	58	18	23	11	12
3	t	63	2	31	32	63
4 rest					10	10
5 rest					10	10
6 rest					10	10
7	s (sphere)	62	10	68	9	48
8	f	49	63	5	13	44
9	E	16	61	33	16	34
10	r	61	62	59	64	59
11 rest					10	10
12 rest					10	10
13 rest					10	10
14	p (planet)	59	9	44	31	5
15	l	55	50	17	44	23
16	a	44	43	43	10	38
17	n	57	0	36	60	28
18	e	48	67	0	3	67
19	t	63	2	31	32	63
20 rest					10	10
21 rest					10	10
22 rest					10	10
23	k (continent)	54	8	66	14	49
24	o	58	18	23	11	12
25	n	57	0	36	60	28
26	t	63	2	31	32	63
27	i	52	41	69	42	61
28	n	57	0	36	60	28
29	e	48	67	0	3	67
30	n	57	0	36	60	28
31	t	63	2	31	32	63
32 rest					10	10
33 rest					10	10
34 rest					10	10
35	k (coastline)	54	8	66	14	49
36	O	26	27	62	52	30
37	s	62	10	68	9	48
38	t	63	2	31	32	63
39	l	55	50	17	44	23
40	I	20	45	2	62	21
41	n	57	0	36	60	28
42 rest					10	10
43 rest					10	10
44 rest					10	10
45	S (shadow)	30	49	49	6	33
46	a	44	43	43	10	38
47	d	47	23	12	28	41
48	O	26	27	62	52	30
49 rest					10	10
50 rest					10	10
51 rest					10	10
52	s (city)	62	10	68	9	48
53	i	52	41	69	42	61
54	t	63	2	31	32	63
55	i	52	41	69	42	61
56 rest					10	10
57 rest					10	10
58 rest					10	10
59	C (chess)	14	64	30	4	22
60	e	48	67	0	3	67
61	s	62	10	68	9	48
62 rest					10	10
63 rest					10	10
64 rest					10	10

Process 5: Cantus firmus to core melody modification

                The resulting cantus firmus is passed through the algorithm "COMPOSE". Here each attribute of each note-event, the pitch, velocity, duration and inter-onset time, is multiplied within a modulus and the result is then added to or subtracted from. The following paragraph gives an example in which only the pitch attribute is affected:

                Imposing a multiplication of 3, a modulus of 12 and a transposition of 60 on the pitch attribute of the cantus firmus results in pitch numbers between MIDI note 60 and MIDI note 72, regardless of the pitch numbers sent into the algorithm. Therefore the resulting pitch numbers are: 60, 63, 66 and 69. This process is shown in Table 2.6.

Table 2.6 Example of possible adjustments made by the COMPOSE algorithm.

Original pitch	Multi-plication	Result	Modulus	Result	Addition	Resulting pitch
23	3	69	12	9	60	69
64	3	192	12	0	60	60
85	3	255	12	3	60	63
46	3	138	12	6	60	66

                Using the table above the speech sound /d/ could be assigned the ID number 47; the pitch 23; the velocity 12; the duration 28 time units; and the inter-onset time 41 time units. The inter-onset time and the duration can be multiplied to create longer time lengths. The same types of adjustments as those imposed on the pitches may be imposed on the other attributes of the cantus firmus. By doing this a melody based in the cantus firmus is produced. Appendix 3 shows the interface for the COMPOSE and CANON algorithms and the process for triggering events.

Process 6: Composing a polyphonic texture from the modified melody

                The resulting modified melody then passes through an algorithm called "CANON", which creates an arpeggio based on the modified melody. The processes this algorithm uses belong to two main types, firstly defining the pitch interval between selected note-events; and secondly defining the inter-onset time between selected note-events. Here the pitch interval process types are listed, beginning with 6.1.1 through to 6.1.7.

                6.1.1: periodic selection of note-events from the modified melody; for example, selecting every third note-event;

                6.1.2: defining the pitch intervals between those selected note-events; for example, if the first selected note is MIDI note 60 and the next selected note is MIDI note 72 then the pitch interval is 12 (72 - 60), if the second note is MIDI note 48 then the pitch interval is -12 (48 - 60);

                6.1.3: multiply the interval numbers by a floating point number or a whole number; for example, an interval of -12 multiplied by 0.25 gives an interval of -3, or if -12 is multiplied by 2 the interval is - 24;

                6.1.4: transpose the resulting interval numbers up or down by adding or subtracting a number from the resulting interval number. For example, if a transposition level of 24 is made, by adding 24 to the interval numbers, then all the interval numbers are raised by 24.

                6.1.5: store the final interval numbers;

                6.1.6: apply the four intervals to the next selected pitch; for example, if the next selected pitch is 60 and the intervals are: -3, 3, 6, and 12, then a chord made up of the MIDI notes, 57 (A), 60 (C), 63 (D#), 66 (F#) and 72 (C) with 60 (C) being the pitch around which the chord is built;

                6.1.7: output the chord pitches at the times specified by the inter-onset time selection process. This creates arpeggios or chords depending on the multiplication of the note-event inter-onset times.

                This process of pitch selection and generating the arpeggio pitches from the selected pitches is repeated four times to produce a four note-event arpeggio or chord. The processes for inter-onset time selection and generating the arpeggio inter-onset time intervals are shown below, starting with 6.2.1. through to 6.2.4.

                6.2.1: periodic selection of note-events from the core melody, as in 6.1.1;

                6.2.2: multiply the inter-onset time of the selected note events by a whole number or a floating point number. This multiplication can be by numbers chosen and preset by the composer, as shown in Figure 2.2 a, or by a ratio between two numbers chosen by the composer, as shown in Figure 2.2 b. For example, if an inter-onset time of 1000 milliseconds is multiplied by 4 it then lasts for 4 seconds, if it is multiplied by 0.25 then the resulting inter-onset time is 250 milliseconds or, if the inter-onset time is multiplied by 0, a chord is produced;

Figure 2.2a Using a preset to set arpeggio delays

Inter onset preset 1 is chosen, resulting in a delay time of 0 msecs. between each note of the arpeggio, that is, a chord.

 

Figure 2.2b Using a ratio to set arpeggio delays
Inter-onset * is chosen. The time is set at 1 then each successive arpeggio delay is multiplied by 0.5.

                6.2.3: store the inter-onset times to be used on the next selected note ;

This creates a polyphony built on the intervals of the elaborated core melody. An example is shown in Figure 2.3a.

                The inter-onset times between notes can be reflected in the generated notes, as shown in Figure 2.3b. The hollow note head is the note generated by "CANON".

Figure 2.3 a and b Example of possible note event simultanaeties (a) and displacements (b) created by CANON.

                The "CANON" algorithm can place up to four pitches alongside a melody pitch. These pitches can create a chord, as in Figure 2.3 a, or can be time displaced in relation to the time displacement of the chosen melody pitches, as in Figure 2.3 b. Here the displacement is similar to the displacement of the chosen pitches, that is, one 1/4 note. This time displacement can be lengthened or shortened to fit the needs of the composer.

                The philosophy behind the "CANON" algorithm is that if the supporting polyphonic and harmonic background of a melody reflects the pitch and time intervals of that melody a cohesive and predictable relationship between the foreground and background is produced. This then provides a holistic and tightly integrated musical event for the listener. Appendix 3 describes how "COMPOSE" and "CANON" are used in the three studies presented here.

Process 7: Playback

                In Studies 1 through to 7, (audio tracks one to seven on the compact disc One), a Korg 05R/W MIDI module is used for playback. The piano sound was chosen because it brings forward the rhythmic attributes of the compositions, due to the very fast attack of the piano sound, and the harmonic attributes by the interplay of sustained strings. Using the Korg 05R/W means that the compositions are not re-interpreted by a musician. Therefore the timing of each note and the velocity with which each key is triggered reflects the computer composition as accurately as possible.

EVALUATION

                The seven studies given here are firmly rooted in the serial, twelve tone, musical traditions. As mentioned above the phoneme string provides a melody that is difficult to reconcile within western musical expectations and traditions, including the serial traditions of the middle and late twentieth century. The sixty-nine possible pitches used here far exceed the range of twelve possible pitch classes commonly used in serial music. This is also the case for the other three elements, amplitude, duration and inter-onset time, used in the pieces.

                Another mitigating factor in relating this music to its roots is that the order of note-events is not repeated as regularly or predictably here as in traditional serial music. The most effective result of this wide range and lack of regular repetition is that the point around which the listener can orient himself or herself is elusive when compared to more traditional musical structures.

                The problem of the listener's orientation is partly overcome by the use of the CANON algorithm. By creating other note streams which reflect the core melody either as a simultaneous, vertical harmony, as in studies two and three, or as an arpeggio, or horizontal harmony, as in study one, a context based on the core melody is provided. This process is most valuable in study two where the full range of the core melody is used.

                Problems resulting from the large range of pitch, velocity, duration and inter-onset times are partly overcome in the COMPOSE algorithm. This is done by simply reducing the range of some of the elements. In each of the studies presented the velocity of the core melody remains intact. This maintains one unique attribute in each of the note-events in the core melody.

                By reducing the range of the note elements used in the core melody, as in studies one and three, there is greater repetition of pitch, duration and inter-onset times. This results in a more predictable rhythmic structure, particularly in study three, where note durations and inter-onset times are based on traditional metric subdivisions.

                The main success of this process is the non-mechanised, almost improvised nature of each piece. This resonates with the improvised nature of speech, even when it is constrained by the stylistic requirements of poetry.

                On compact disc One there is the full version of ZOOMING IN and extracts of music based on two other poems which have been played through the same algorithms. These two poems are Ambit, again by Alex Skovron and Not Yet One by another Australian poet, Earl Livings; they are given, in full, in Appendix 5. Only the first stanza of Ambit is used in the example here. These are provided as examples of how this process of creating music from a text works with other texts. In each case exactly the same algorithms are used.

                Areas for expanding this process and exploring it further include testing it with poems which use more predictable rhyme and rhythm schemes, such as limericks. This is somewhat explored in Ambit and Not Yet One, both of which have more predictable rhyme schemes than ZOOMING IN. Doing this could create more repetition and redundancy in the core melody resulting in the listener orienting themselves more easily.