Trial_Queries.txt

Trial Queries:
==============
1.) Compare the frequencies of interval n-grams in a mass and its contrapuntal model.
2.) Plot contrapuntal entropy against composer (or any metadata field).
3.) Plot length of n-gram vs. frequencies in a set, coloured according to piece they occur in (scatter plot)
4.) Plot intervals against beatStrength (i.e., "how often does a dissonant interval occur on a strong beat?")
5.) Compute the dissonance level of a set of pieces, plot on bar graph. "Aggregators. Dimensional reduction."
6.) Given a set of pieces, divided into a set of categories (i.e., 50-year chunks), be able to treat each category as though it were its own piece, to run through all the other analyses ("categorize datasets")
7.) Filter n-grams so that n-grams that are part of a larger-value-of-n-gram don't appear in the smaller list.
8.) Calculate frequency of Roman numerals against year.
9.) Create statistical models. Given a sample set, create a Markov transition matrix, for example. Calculate n-gram transition probabilities. Includes "smoothing/priors," "balances vs. proportions," etc.
10.) Score of a piece showing all the k-voice contrapuntal modules of a length >= n that also appear in another piece.

VisTrails Modules We'll Need:
=============================
- name: description
    - inputs: ?
    - outputs: ?
    - parameters: ?
- piece
    - inputs: list of pathnames
    - outputs: list of IndexedPiece
    - parameters: (optional) which indices to create
        - if no indices specified, default is pitch/rest/chord (i.e., all the vis9c indices)
        - we "have to" load pieces now, so we can get metadata; since we load them, might as well index too
- make k-part simultaneity (including "make intervals" shortcut for 2-part simultaneity)
    - inputs: list of IndexedPiece
    - outputs: list of IndexedPiece
        - adds an index of simultaneities according to the parameters
    - parameters: simultaneity of what, "repeat identical," "offset between events," and list of the parts to include (probably as list of int)
- make n-grams
    - inputs: list of IndexedPiece
    - outputs: ?
    - parameters: thing of which to make n-grams (i.e., "interval," "chord," etc.), value of 'n', "repeat identical," "offset between events," and list of parts
        - these are all required, so it can choose the right index from the IndexedPiece; if it doesn't exist yet, it'll be created, so in the VisTrails you don't have to use "make k-part simultaneity" unless you want
        - sometimes we should be able to guess "thing of which to make n-grams" (i.e., when there is only one index; when the previous module was "k-part simultaneity")
- choose a metadatum
    Extract a metadata field from all the pieces.
    - inputs: list of IndexedPiece
    - outputs: dict of string
        {'field': 'whatever field', 'pathname1': 'value', 'pathname2': 'value'}
        --> 'field' is the "the metadata field to choose" parameter
    - parameters: the metadata field to choose
- categorize
    Use a metadata field, or more than one, to create named categories for comparison
    - inputs: list of IndexedPiece (that comes out of "piece")
    - outputs: dict of list of pathnames
        e.g., {'Beethoven': ['pathname1', 'pathname2'], 'Schubert': ['pathname3', 'pathname4']}
        e.g., {'mass': ['kyrie.krn', 'gloria.krn'], 'model': ['Lhomme_armé.krn']}
    - parameters: output(s) from "choose a metadatum" above
- make n-gram transition model
    - inputs: ?
    - outputs: ?
    - parameters: ?
- calculate entropy
    - inputs: ?
    - outputs: ?
    - parameters: ?
- observe n-gram 'n'
    - inputs: ?
    - outputs: ?
    - parameters: ?
- "repeat" sub-workflow (already in VisTrails?)
    - inputs: ?
    - outputs: ?
    - parameters: ?
- beatStrength
    Use an algorithm to find the beatStrength of given offsets. For now, the only algorithm is "music21," which just uses the beatStrength property of the music21 object
    - inputs: list of IndexedPiece
    - outputs: list of IndexedPiece
        - adds an index of beatStrength
    - parameters: "offset between events," algorithm to use for finding beatStrength
- merge (e.g., intervals-and-beatStrength)
    Combine results across categories.
    - inputs: ?
    - outputs: ?
    - parameters: which category combinations to make lists of, some way to know which index/aggregation to use
        - could be trivial, but in cases with lots of categories, it'll be really important to know the user only wants two lists: one with things that happen in categories A and B, and one with things that happen in categories C and D
- mean
    - inputs: ?
    - outputs: ?
    - parameters: ?
- remove embedded n-grams
    - inputs: ?
    - outputs: ?
    - parameters: ?
- find chords in polyphonic textures
    - inputs: list of IndexedPiece
    - outputs: list of IndexedPiece
        - adds an index of chords, with some way to know they're "reduced chords"
    - parameters: which parts to consider (default to "all"), longest possible chord (i.e., how far to search for a complete sonority)
- label chords
    Given an algorithm with which to label chords, do it. We'll provide some defaults: RN, FB, RN/FB
    - inputs: list of IndexedPiece
    - outputs: list of IndexedPiece
        - adds an index of chord labels
    - parameters: which chord-labelling algorithm
- balances
    - inputs: ?
    - outputs: ?
    - parameters: ?
- filter (e.g., "if n-grams appear in all categories" or "if more than one occurrence")
    - inputs: ?
    - outputs: ?
    - parameters: ?
- LilyPond (worry about that later)
    - inputs: ?
    - outputs: ?
    - parameters: ?