How to annotate musical events in kern files

[WolfgangDrescher]

I started a discussion on how to label cadences and cantus firmi in #779. Maybe we can take this a step further and find a general solution for this.

I am thinking of a (or multiple) **annot spines to annotate such things. Probably it is a good idea to keep data records of this spine relatively open for all kind of comments. But at the same time there should be a list with some standardized annotations so they can be used across projects.

Here is a list of some musical events that happen regularly enough to discuss about a standardized encoding:

1. Cadences
2. Clausulae
3. Sequences
4. Marking Cantus firmus (mainly renaissance music)
5. Leitmotifs or quotes from other pieces
6. General comments
7. Figures (such as Superjectio, Anticipatio, Transitus irregularis, etc.)

It would be nice to have an additional filter that allows to display labels for the annotations as !LO:TX or similar. E.g. "Quintfallsequenz" could be a good label to display in the score when annotated in an **annot spine. Ideally this would automatically mark the segment as a phrase with { and }. It should be possible to dynamically show and hide labels and phrase markers in verovio.

This could be similar to what @craigsapp wrote in https://doc.verovio.humdrum.org/humdrum/scale_degrees/:

Contours across slur/phrase boundaries
In a future extension to the deg command, phrases ({}), slurs ((), and/or fermatas (;) can be encoded in **deg data in order to disallow labeling melodic contours to cross them.

[WolfgangDrescher]

Cadences

In my project im currently using global comments:

!! cadence start b
2.A	spot-	2c	ten/	2.cc	spot-
*	*	*	*Xij	*	*
.	.	2e	zu	.	.
4G	.	.	.	4b	.
1F	.	4f	spot-	1a	.
.	.	4c	.	.	.
.	.	2f	.	.	.
=28	=28	=28	=28	=28	=28
0cll	ten.	0ell	ten.	0gll	ten.
!! cadence end b; ultima=c

https://github.com/WolfgangDrescher/lassus-geistliche-psalmen/blob/f9ab05de13ea119c4e7d4d33d96e45176aa3b7c3/kern/01-beatus-vir.krn#L199-L210

Note that cadences need to be labeled when they are overlapping. Here is an example for an overlap:

https://github.com/WolfgangDrescher/lassus-geistliche-psalmen/blob/f9ab05de13ea119c4e7d4d33d96e45176aa3b7c3/kern/01-beatus-vir.krn#L192-L210

---

With an **annot spine this example could look like this:

VHV

**kern	**text	**kern	**text	**kern	**text	**annot
*staff3	*staff3	*staff2	*staff2	*staff1	*staff1	*
*clefC3	*	*clefC2	*	*clefG2	*	*
*k[]	*	*k[]	*	*k[]	*	*
*C:ion	*	*C:ion	*	*C:ion	*	*
*M2/1	*	*M2/1	*	*M2/1	*	*
*met(C|)	*	*met(C|)	*	*met(C|)	*	*
*MM180	*	*MM180	*	*MM180	*	*
2.A	spot-	2c	ten/	2.cc	spot-	cadence start a
.	.	2e	zu	.	.	.
4G	.	.	.	4b	.	.
1F	.	4f	spot-	1a	.	.
.	.	4c	.	.	.	.
.	.	2f	.	.	.	.
=28	=28	=28	=28	=28	=28	=
0cll	ten.	0ell	ten.	0gll	ten.	cadence end a; ultima=c
==	==	==	==	==	==	==
*-	*-	*-	*-	*-	*-	*-

Or even with the new **deg spine additionally to get rid of ultima=c:

VHV

**kern	**text	**kern	**text	**kern	**text	**annot	**deg
*staff3	*staff3	*staff2	*staff2	*staff1	*staff1	*	*
*clefC3	*	*clefC2	*	*clefG2	*	*	*
*k[]	*	*k[]	*	*k[]	*	*	*
*C:ion	*	*C:ion	*	*C:ion	*	*	*C:ion
*M2/1	*	*M2/1	*	*M2/1	*	*	*
*met(C|)	*	*met(C|)	*	*met(C|)	*	*	*
*MM180	*	*MM180	*	*MM180	*	*	*
*	*	*	*	*	*	*	*hide
2.A	spot-	2c	ten/	2.cc	spot-	cadence start a	.
.	.	2e	zu	.	.	.	.
4G	.	.	.	4b	.	.	.
1F	.	4f	spot-	1a	.	.	.
.	.	4c	.	.	.	.	.
.	.	2f	.	.	.	.	.
=28	=28	=28	=28	=28	=28	=	=
0cll	ten.	0ell	ten.	0gll	ten.	cadence end a	1
==	==	==	==	==	==	==	==
*-	*-	*-	*-	*-	*-	*-	*-

---

• Probably the ordering for **annot data tokens is better the other way around as start cadence and then add additional parameters with a semicolon as a separator.
• Maybe the label for overlapping cadences should be an attribute such as: start cadence; key=a.
• An alternative for ultima=c could be deg=1 or something like scale-deg=1. Although the key interpretations would have to be added to the **annot spine.

[WolfgangDrescher]

It is to be discusses whether additional parameter should be added to the starting or to the ending record of a cadence. For ultima=c I thought that it suits well to the end because it is next to the corresponding note. Maybe for one annotation the parameters of the start and end record could be collected and combined where the parameters of the end record overwrite the ones from the start record.

[WolfgangDrescher]

Sequences

Here is a document that I usually use for teaching sequences: Satzmodelle und harmonische Wendungen.pdf

Possible encodings for **annot data records could be:

• start sequence; type=descending-fifths; label=Quintfallsequenz
• end sequence
• start sequence; type=fauxbourdon; label=7-6-consecutive
• start sequence; key=a; type=romanesca
• end sequence; key=a
• start sequence; type=diatonic-carousel; label=Karussell
• etc.

Where the label could be displayed like !LO:TX. Especially for sequences showing phrase markers will be beneficial.

Names for the types are to be discussed as the ones I used in my handout are probably biased with German terminology. E.g. Karussell or Teufelsmühle (see also Voglerscher Tonkreis or Omnibus). But of course terminology used in the US like the Prinner should also be possible.

Here is a list with sequences that I would like to annotate:

• Fauxbourdon
• Lamento (diatonic and chromatic; with and without 7-6 suspensions)
• Gymel (parallel 3rds or 6ths)
• Romanesca (with and without 4-3-9-8 suspensions)
• Karussell (diatonic and modulating)
• Quintfall
• Quintanstieg
• Fonte
• Monte
• Ponte
• Teufelsmühle / Voglerscher Tonkreis / Omnibus

I will add items to this list then something new comes to my mind.

[WolfgangDrescher]

Clausulae

When doing analysis on paper I usually would simply write SK (Sopranklausel), TK (Tenorklausel) and BK (Bassklausel) to the right of the ending note of a clausula. So when encoding this it is not clear where to mark the start of a clausula.

E.g. a Sopranklausel (cantizans) could mean a scale degree progression 1-7-1 or (more typically) simply 7-1. And it gets more complicated when there are ornaments.

This is why I suggest to have a start and an end record. But maybe we could make the start record optional.

In general it's easy to automatically calculate the clausulae when the cadence end is marked (check this function; could probably be improved with deg). So it will probably be sufficient to only encode more complex clausulae.

Possible encodings for **annot data records could be:

• start clausula; type=cantizans
• end clausula
• or maybe also start cantizans / end cantizans
• maybe parameters like label=SK; label-position=end
• and label=cantizans; label-position=start

Here is a list with clausulae names that I would like to annotate:

• cantizans
• altizans
• tenorizans
• bassizans

How should it be encoded when a clausula is evaded (in all different ways: stepwise in the wrong direction, jumps, resolves into a rest)?

Also check this screenshot I took from the Video about cadences of the Early Music Sources YouTube channel:

Source: https://youtu.be/jaCRUdxTRSM

Combining annotations for clausulae with annotations for ultima, penultima, antepenultima will be useful. We could allow those values as data records between start clausula and end clausula.

[WolfgangDrescher]

Note that for cadences and sequences the **annot spine was marking a slice globally for all voices while for clausulae it should be bound locally to a **kern spine. Maybe something like **lannot or **annotl will be a good idea?

[WolfgangDrescher]

Cantus firmus

Currently I am using !LO:TX:a:i:t=c.f. to display the c.f.. But I think this is semantically not correct as it is not part of the original document. Check out an example here.

This is why I would prefer to use !cf and transform it into !LO:TX:a:i:t=c.f. with shed. But probably this is also a good annotation that could be added to **annot.

Possible encodings for **annot data records could be:

• start cantus firmus
• end cantus firmus
• start cantus firmus; label=c.f.

!LO:TX:a:i:t=c.f. could automatically be added by the start cantus firmus record.

Note that end cantus firmus can be made optional but it is needed since there are examples where the cantus firmus is wandering through the voices. Or sometimes just a few notes are taken from the cantus firmus but after it continues freely.

Compare https://www.digitale-sammlungen.de/de/view/bsb00075346?page=136 (Tricinium by Lasso) with https://www.digitale-sammlungen.de/de/view/bsb11116252?page=87 (composed by Ulenberg and used as c.f. by Lasso).

For this exmaple it would be nice to somehow show the end of the cantus firmus. Either with phrase markers or with a label such as end cantus firmus; label=c.f. end.

[WolfgangDrescher]

Note that again **annot would need to be bound locally to a **kern spine.

[WolfgangDrescher]

Humlib filter (for controlling the score in verovio)

It would be nice to have an additional filter that allows to display labels for the annotations as !LO:TX or similar.

This filter could also show and hide any of the annotations.

Possible functionality of this filter could be:

• Show all cadences and display them with phrase marker
• Hide all clausulae and show label and phrase marker
• Show all clausulae but only the label
• Show all annotations
• Hide all annotations
• Show all descending fifths sequences with label and phrase marker but hide all other sequences
• Show all sequences but only with phrase marker and without the labels
• etc.