Irena
Mitevska
Problems in the Digitization of the
Firfov Collection of
Transcriptions
The
Firfov Collection of Transcriptions was created in the 1970s. The Firfov family,
who gave us this collection, assumes that most of these songs have not been
published. IRAMÕs archivation of this collection had the primary purpose of
forming a digital database with the collectionsÕ transcriptions, only to
continue to create a secondary database containing data about the content of
the collection, and finally to publish the unpublished material. (Mitevska,
2004).
The
secondary database was completed in June, 2004. It consists of 14 fields formed
according to the data categories established by Firfov. There was no other
source of secondary data apart from the data left by the very transcribers.
Firfov left lots of information for some of the transcriptions, but there were
such where some basic data such as title, lyrics, tempo and informer were
missing or incomplete. The distribution of data in the secondary database of
the collection was thoroughly presented in one of our previous papers
(Mitevska, Irena. 2004. The Distribution of Data in the Firfov Transcription
Collection; Contemporary Trends in Musicology and Ethnomusicology, pp.52:64, Skopje: IRAM).
The
digitazation of music manuscripts should start with scening of the originals.
The scaning of The
Firfov Collection of Transcriptions was done with EPSON perfection 610 scener, with
the following preferences:
- mode: profetional mode
- document type: refletive
- document source: document table
- auto exposure type: photo
- 24 Ð bit color
- resolution: 300 dpi
- document size: width 8.5; height
11.7
- target size: width 8.5; height
11.7
The Firfov Collection of
Transcriptions was
digitized with the Sibelius 1.4 software. The digitization of the transcriptions brought into light
many problems regarding the entering of the notation and the lyrics, the lack
of data, inconsistency in some of the existing data, etc. The digitization of
the The Firfov Collection of Transcriptions was completed in January, 2005.
After
the completion of the digitization, we directed ourselves to comparing the
already published transcriptions by Zhivko Firfov with the transcriptions from
our collection, also comparing the digitized transcriptions with the digitized
audio recordings from the Audio Collection Firfov.
This
paper discusses the problems we encountered when digitizing this collection and
compares the already published transcriptions and the Firfov Collection audio
materials. Thus, this paper consists of two parts:
I. Problems regarding the digitization of the The Firfov Collection
of Transcriptions.
II. Comparison of the The Firfov Collection of Transcriptions with published, digitized material.
I. Problems regarding the digitization of the The
Firfov Collection of Transcriptions
The
problems arising from the digitization of the collection are classified as
follows:
1. Problems in the digitization of the notation
2. Problems in the digitization of the lyrics
1. Problems in the digitization of the notation
1.1. Illegible notation
During
the digitization of the The Firfov Collection of Transcriptions, the problem of an unclear tune
arose numerous times. Since we were left with the original manuscripts, in some
parts of the tune we could not discern the appropriate tone. For instance, in
the following transcription it is not clear whether the tone in question is ÔgÕ
or ÔaÕ.
1.2 Lack of data
While
designing the secondary database we realized that part of the transcriptions lack
various data such as: lack of meter notation, lack of title, lack of lyrics,
incomplete lyrics and tune. The lacking data was divided into two groups:
-
missing
titles, lyrics and tunes
-
missing
meter
1.2.1 Missing titles, lyrics and tunes
During
the digitization of the The Firfov Collection of Transcriptions there emerged examples were key
transcriptional data was missing, such as lack of the title of the song or of
the lyrics (picture 1). We encountered situations were the lyrics or the tune
were not fully noted, i.e. only one stanza or even verse from the lyrics was
noted. There were also examples were it was clear that the end of the tune was
also missing or data regarding the instrumental intro. Picture 2 shows that it
is unclear whether the written tune is an instrumental intro or the lyrics are
missing. The completion and reconstruction of the missing transcriptional
sections (the song lyrics, titles or the entire tune) is undoubtedly a very
complex and compelling task which requires further research. The complexity of
the task also arises from the multiple variations of one folk song. In fact,
these variations depend on many issues, including the different interpretations
by one or more informers and the conditions of recording.
picture 1
picture 2
1.2.2. Missing meter
Most
of the transcriptions involved problems regarding the meter. In some instances,
the beginning noted one meter, but the bars clearly showed that the
transcription is in another meter, for example:
-
the
beginning notes a 7/8 meter, but the original rhythmic values prove that the
meter is in fact 7/16.
-
the
beginning notes a 7/8 meter, whereas the following bars are 8/8.
We
encountered examples were the meter hadnÕt been noted at all. These instances
were corrected during digitization, as it was clear that this was the
transcriberÕs error.
In
addition to this, the The Firfov Collection of Transcriptions included examples where the tuneÕs
metric changes hadnÕt been noted. These changes regarded one or more bars. In
such cases we were not certain whether the transcriberÕs error was in question,
or the entering of the changes of the meter.
The
Sibelius
software used for digitization includes an option irregular (create/bars/irregular) which
enables the entrance of irregular bars, i.e. bars in other meters. Thus, the
transcriptions were digitized as the transcriber originally transcribed them,
leaving room for further changes.
Apart
from such cases, we encountered examples where the beginning of the
transcription notes two meters, for instance 2/4 and immediately thereafter
3/4.
It was obvious from the bars that followed that in these examples the
meter alternated with each bar. However, the meter cannot be entered in the Sibelius program in such a way. We therefore
solved the problem by noting each meter separately when changes occurred.
1.3. Noting several versions of parts of the songs
In
most of the transcriptions noted as monophonic tunes there were instances where
a second voice appeared only for one or two bars. We concluded that these cases
are another version of that part of the tune because of the change in the
lyrics in some of the stanzas. It would unquestionably be obscure to note these
instances as a second voice appearing only in one part of the transcription. We
decided to enter these examples through one of the Sibelius options Ð ossia above/below in the
menu Create/extra staff. This option enables entering of another staff with as
many bars as necessary in the other version above or under the problematic
staff.
(picture)
1.4. Entering the transcriptions without meter
In
The Firfov Collection of Transcriptions there were examples which did not include
meters, i.e. examples performed freely. Again we used the irregular option and we entered these examples
bar by bar.
1.5. Entering the accidentals before the staff
Part
of the transcriptions contained an unusual combination of accidentals noted in
the beginning of the staff, for instance ÒF-sharpÓ and ÒE-flatÓ.
We presume that such cases suppose a harmonic minor. Since the Sibelius program does not offer such an
option, we solved the problem by alterating the separate tones in the course of
the tune.
2. Problems in the digitization of the lyrics
2.1. Illegible handwriting
During
the process of digitization we often encountered parts from the lyrics (words)
that had been illegibly written. Some of the unclear lyrics were not
understandable because we used photocopies for the digitization, and not the
original transcriptions. When comparing the photocopies to the originals, some
of the unreadable words were cleared. The remaining parts were unclear owing to
illegible handwriting; those were not entered and were marked by question
marks.
2.3. Inconsistency
between the lyrics below the tune and the written lyrics
The
digitization of the collection also raised the issue of the inconsistency
between the lyrics below the tune and the separately written lyrics. Although
these lyrics were identical in most of the transcriptions, there were examples
where some parts differed.
Usually
the changes between the lyrics were minimal, often only one or two words. For
example, the name mentioned in the lyrics differs (ÒVidoÓ below the tune, and
in the separately written lyrics Ð ÒBojeÓ).
Since
we considered it important to include all the details from the lyrics just as
they had been noted by the very transcriber, when digitizing we decided to note
the different words in parentheses. If there was an entire differing stanza, we
presented it parallely with the stanza written below the tune.
2.4. The authorÕs inconsistency in noting
Digitization
of the lyrics revealed examples where before the verses there were numbers with
illogical patterns. Usually these were transcriptions that appeared several
times in the collection. Therefore, we assumed that the informer improvised in
the lyrics and that is how the lyrics were noted. In such cases we copied the
lyrics from the identical complete transcription in the collection.
There
were examples in some of the lyrics where the separately written lyrics lacked
the additional words, such as lele and dzhanam[*]. We did not consider this a
problem, since the tune implies their position in all stanzas.
2.5. Repetition of verses
In
most of the lyrics, some verses are repeated several times. Thus, for instance,
the transcriber used the sign Òx2Ó under the first stanza, but in the remaining
stanzas this sign is absent. There is a similar example with noted repetitions,
even though the tune clearly shows that part of the stanza will not be
repeated. There were reverse instances, where no repetitions were noted in the
melody, but the lyrics clearly indicated that there should be repetitions.
In
both cases the lyrics were noted in accordance with the original, without any
acts of intervening.
2.6. Noting parts of the lyrics where the transcriber was uncertain
The
transcriber was uncertain regarding some parts, so he wrote question marks next
to the stanza or verse. Also, he was not certain of some words which he completely
omitted, leaving question marks in their positions. There examples were also
noted without corrections, as they were originally written by the
transcriber.
II. Comparing The Firfov Collection of
Transcriptions with the remaining material
After
the digitization of The Firfov Collection of Transcriptions was completed, our next step was to
compare the already published material by Zhivko Firfov with our collection in
order to establish the number of published transcriptions. Another comparison
was also made with the secondary database from the Audio Firfov Collection digitized at IRAM, 2000, in order
to see how many of the transcribed songs have audio recordings.
1. Comparing the secondary database from The Firfov Collection of
Transcriptions
with the publications
In
order to establish whether some of the transcriptions in the The Firfov
Collection of Transcriptions had already been published, we compared this collection to the three
published collections of Macedonian folk music songs containing transcriptions
by Zhivko Fifov: ÒMacedonian Music Folklore Ð songs - 1Ó, ÒMacedonian Music
Folklore Ð songs - 2Ó and ÒSokol mi leta visokoÓ (The Falcon Flies High).
ÒMacedonian Music Folklore Ð songs - 2Ó and ÒThe Falcon Flies HighÓ contain
transcriptions exclusively by Zhivko Firfov, whereas ÒMacedonian Music Folklore
Ð songs Ð 1Ó also includes transcriptions by other authors.
We
compared the titles of the songs in these collections and the titles from the
secondary database from The Firfov Collection of Transcriptions. This comparison revealed 85
transcriptions with identical or similar titles.
We
then compared the tunes and lyrics from the transcriptions with similar titles.
In 14 separate cases, i.e. transcriptions, the song had the same title, but it
was nevertheless a different one.
After
this elimination, we found examples with the other transcriptions that differed
in the tune and lyrics, implying that these are different versions of the
songs, even when the songs had identical lyrics and different tunes. In the
difference in tunes there were examples of:
-
identical
lyrics and tune, but different meter
-
identical
tune, but different tonality
-
identical
lyrics, but completely different tune
This
comparison concluded that only 19 transcriptions are identical with the
transcriptions from the collections, whereas the remaining 55 are different
versions.
2. Comparison of the secondary database from The Firfov Collection
of Transcriptions with the secondary database from the Audio Firfov Collection
Among
the assumptions regarding the appearance of the Firfov Collection is that Zhivko Firfov recorded this
material with transcriptional purposes. We thus expected more examples
contained in both collections, and even some earlier published songs. On the other
hand, in 1959 Firfov himself believed that:
ÒThis collection
should mark the beginning of a new practice of publishing folk songs here [in
Macedonia]. So far we have not been audio documenting the published material.
Today, in the times of mechanized musicological action, this practice should be
completely abandoned. We would thus avoid many older and recent
misunderstandings regarding different questions which concern the features of
our music folklore.Ó (Firfov, 1959)
The first step of comparison was to
separate the fields from both databases which would most easily and quickly
give us identical examples. Those were the fields ÒtitleÓ and Òtext incipitÓ
used as categories in both databases. After separating the fields we needed, we
put them in a table in the Microsoft Excel software.
The
next step of comparison involved the use of the tool ÒsortÓ. We sorted the data
in alphabetical order in order to see if there were identical or similar titles
and lyrics from the examples in both collections.
After
the sorting, 96 identical or similar titles emerged. Having compared the text
incipit of the examples with similar titles, we found that 29 transcriptions
were versions of the lyrics. We marked these examples as ÒpossibleÓ identical
songs.
The
following step will be re-listening to the entire audio material before finally
concluding whether the songs are identical, different, or versions.
Perspectives
Apart
from archiving this material, the digitization of The Firfov Collection of
Transcriptions opens
other perspectives for further theoretical and other types of analyses.
At
first, the fact that this material is in digital format will enable it to be
listened through MIDI performance. Although the MIDI performances of the music
folklore transcriptions essentially differ from the real performances of the
folk material, in absence of the opportunity of the traditional notational
system to present the ornaments, rhythmic and agogic deviations, as well as the
specific tone series which by rule contain micro tones, this hearing will
present the first picture of the transcribed material. It will also separate
some errors and illogical data in the transcribed material, leaving room for
further correction.
An
even more precise picture of the relations between the transcribed material and
the audio from the Firfov Collection is expected from spectral analysis.
In
fact, these are our next steps in working and analyzing the already digitized
material at IRAM.