The text corpora originally organized by Prof. Mark Davies at BYU now have their own website.
https://www.english-corpora.org/overview.aspUnfortunately, I believe it no longer offers access via the old user interface, which was easier to use and, very importantly, did not require JavaScript. Now that it no longer works without JS, I seldom use the website myself; ditto the Uni Leipzig corpora, which have fewer features but many languages.
Nevertheless, if you take some time to read all of Mark Davies's instructions and examples (not at all easy to find from the now very bland and empty homepage, but they should be there somewhere), I believe there are still many ways to customize a search using parts of speech or even other descriptors. For instance, I believe you could probably create a search on
"("song" OR "album")+("by" OR "from")+({person} OR {name})"
and it would be pretty much the results you want. I don't know the exact syntax and punctuation you would have to use, but you could figure it out.
The other choice is just to ask us native speakers, and we can just tell you that 'by' is the default, always correct and usually preferred.
In some cases 'from' is also possible, if there is some concept of movement, like a song just released, which is considered to move
from the artist out
to the public, so that it
comes from the artist, with the verb 'come' often understood at some level.
