Encoding relationships between annotations

[jonnybluesman]

In the same JAMS, there might be annotations (with same or different namespace) that refer to other annotations. We should keep track of these relationships in the sandbox, or somewhere more appropriate. This also requires an identifier (in the JAMS file) for each annotation. Examples of possible relationships:

• Same as -- when 2 annotations are basically the same, but they are expressed in different ways.
• Converts -- when 1 annotation is the output of a conversion / simplification / pre-processing.
• Extends or derived from -- when 1 annotation extends another, or is simply derived from another annotation (e.g. a measure of chord complexity derived from a chord progression).

To some extent, there are cases when some of them can be inferred from the annotation metadata (e.g. "Jacopo" has produced 2 chord-harte annotations of the same piece, so, these can be the same?), but preferably, we should make things explicit.

[jonnybluesman]

Another use case

This is also useful when we have multiple harmonic annotations for the same piece in the same dataset. For example, in Schubert Winterreise we have: 1 global key in the (score / audio) metadata; and several chord and key annotations produced by different annotators. Here, to read 1 chord annotation, we cannot refer to any other key annotation; but we need to make sure that we retrieve the key_mode annotation produced by the same annotator. Another striking example is the CASD.

This may call for another type of relationship between 2 annotations; something like related_to.

[jonnybluesman]

As JAMS cannot handle this natively, we are moving this to define the requirements of JAMS-X.