OntologyDesign sprint: iteration 1 (places)

jonnybluesman commented 3 years ago

As @enridaga suggested, at the current iteration we are focusing on basic concepts + those related to places. Together with @andreamust and @valecarriero, we have filtered the place-related "fields" (from our list [1]) and tried to contextualise them with the current (and the potential) design of the ontology modules.

In sum, these are the data types related to places, with the corresponding entries in [1]:

Place mentioned / subject of the composition, e.g. "Penny Lane, Liverpool, UK" in "Penny Lane" by The Beatles (10);
Facts / curiosities related to the latter, e.g. "In Penny Lane, there is a barber showing photographs..." (11);
Country of each of the artists of a composition, e.g. "GB" (45);
Geographical area where each artist (of a composition) has started their career (46);
Recording place (studio, live venue) of the recording, e.g. "Abbey Road Studios: Studio 2" (65);
Address of the recording place (related to the latter), e.g. "3 Abbey Road, St John's Wood, London" (66);
Places where the release (album) was "made official", e.g. US, EU (75).

Please note that this is more complex from a technical point of view, as MusicBrainz is very granular and provides information at different levels. Indeed, a musical work (the composition) can have N recordings, each of which can be present in M releases. For data collection, we tried to simplify this step, so we can assume a 1:1:1 relationship for the moment.

Anyways, and more importantly, we were wondering if all these data types should be modelled in our ontology modules.

[1] https://liveunibo.sharepoint.com/:x:/r/sites/polifonia/Shared%20Documents/WP6/AI%26Music@Sonar2021/datasets/overview.xlsx?d=w352ea4d8aefe40cbb080d1c291c52998&csf=1&web=1&e=V2BAIm

enridaga commented 3 years ago

Anyways, and more importantly, we were wondering if all these data types should be modelled in our ontology modules. I think they all should be modelled, not sure if the modelling should be done with typing, probably not. But this is a question for the ontology engineering team.

An important question is which ones of these can be added to the datasets for the demo ASAP. (See also #10)

jonnybluesman commented 3 years ago

The datasets where all these fields come from are already in the datasets branch, but we can re-organise the structure of the repository to make it more readable and easy to navigate. From an analysis of the current data, we can conclude that:

Places 1 and 2 are only available for 5 tracks from The Beatles and just 1 from Queen.
Places 3 and 4 are available for every single artist that contributed to the composition (generally available).
Places 5 and 6 are related to the recording and are given for "mixed at" and "recording at" (generally available).
Place 7 is not very informative apparently (e.g. EU, XE are too general) and is associated to all the possible releases (albums) where the recording appears.

In sum, we have all the data but we can rely mostly on 3, 4, 5, 6. As for the ontology modules, we are constantly in contact with @valecarriero, so our concern in this issue was actually shared.

jonnybluesman commented 3 years ago

Some modelling progress from the OD team curating the musical performance module (@valecarriero, @jonnybluesman, @andreamust). This partial model generalises some key parts of MusicBrainz (where we get most of the data), is still aligned with our competency questions, and addresses Places 5, and 6 (see list above).

recording-v2