I look forward to the new search algorithm. Let us know what changes if any we need to make to our .ly files. Presumably I would need to index the variants as individual search sections.

Sorry about your car. Let's hope the New Year is more peaceful for us all.

I haven't addressed this issue lately because I am working on a better search method that can find a theme anywhere in a piece, not only if the query begins at the same point where the stored melody begins. That will make it possible to include - and find - multiple themes in one Lilypond file, so this issue should be automatically solved as a side effect.

The new search method will take into consideration rhythm and exact intervals, but it will still tolerate deviations (the distance measure will be continuous, that is, if you get the query just a little bit wrong, the search result will also be just a little bit wrong and you should still find the result you are looking for). Of course, this method won't work with Parsons Code. I still intend to leave the old method in place as an additional option. The new keyboard (see that thread) is intended for the new search method.

Unfortunately, today I was busy dealing with the mess caused by some (presumably) drunken Dutch people with anti-German feelings. On normal days, they just spit on my car, but yesterday night they set off fireworks on it and smashed a window, so I had to go to the police and deal with my insurance instead of implementing nice new search algorithms.

I have a similar question, so I didn't want to open a new thread.
What is the preferred method of entering multiple melodies for the same song? As Carl pointed out, there are legitimate variants even to tunes as simple as Mary had a Little Lamb. This is especially true for folk tunes, where sometimes the chord structure and rythm is all that remains constant between variants. Would it be best to have separate entries for different variants? Would it be better to notate the variants some other way? In the .ly perhaps?

Music of its nature is sequential therefore in a multi-theme work Theme 1 is followed by Theme 2 Theme 3, etc. Cannot theme numbers be used to get things into an order?

You might well say that a person adding a new melody may not know whether it is theme 1, 2 or 3 or whatever. I say, though, that this does not matter as the editing facility enables adjustments or corrections down the line.

Assume that the theme system is in operation: As an example - assume Theme 1, Theme 2 of an opus already exist in Musipedia. You want to add a melody that comes after Theme 1 but before Theme 2. How? Give the melody a Theme number of 1.5 and it is obvious where it fits. By using numbers and/or decimal fractions to list themes would bring order out of potential chaos. Such a numeric system would also encourage supplementary entries, for example, theme and variations, mottos, or fugue themes.

Existing titles, such as "intro", "chorus", "verse", "famous fragment", could be moved to text box area.

Well, for the verse/refrain of a hymn case, I would want to have only a single piece of sheet music but still have two Parsons codes. (I guess another way to do it would be to enter the Parsons code for the whole song, and then allow searching "within" Parsons codes; but people were worried that this would allow too many false hits. Having two Parsons codes instead is like marking two places in the song as "These are the two places I think people are likely to search for.")

either way, it would be nice if it were possible to add info to the parsons code, such as "intro", "chorus", "verse", "famous fragment", etc.

btw, alternative melodies could, in turn, also have parsons codes for "chorus", "verse", "famous fragment", etc...

so the first solution (bundling) might indeed be more practical.

Sorry about my late response - just came back from ISMIR 2005 (see Paul Lamere's blog: http://blogs.sun.com/roller/page/plamere/20050913).

I think you have a point. Right now, we have a collection of melodies or themes, but no way of expressing that some of them belong to the same musical work. This is a bit annoying since it involves entering metadata such as title, composer, and lyrics multiple times.

I am not sure yet what is the best way of improving it - either make it possible to bundle entries together to express that they belong to the same piece, or make it possible to add multiple melodies to one entry (that would mean, multiple Parsons codes, multiple bits of sheet music, and also multiple fields for metadata).
The first solution could involve a new "works" table with some metadata for the whole work and an ID field. The existing melody records would then just need one workID field to hold the information which work they belong to.
I think I'll do it like that unless someone else here comes up with a better solution.

Cheers,

Rainer

I think it would be good if you could have multiple Parsons codes per entry. I can think of at least two uses for such a thing:

I tried to look up "Mary Had a Little Lamb". The version in your database is not the same one I learned; but I hesitate to "fix" it, because it's easy to believe that the tune in the database is a legitimate variant.

For many songs (I'm thinking particularly of hymns), it would be good to index multiple sections (for instance, the verse and the refrain) separately -- sometimes I can remember the refrain but not the main verse, sometimes vice versa.