FR24 published their ADS-B data in csv format.
Airspeed isn't there, but GPS lat/long is, and that's what I used to get a ground speed of 300 kts until nose down. Even that data is weird, though,since not only does the same data repeat, sometimes it throws other numbers in there.
So I found that I could get acceptable data by taking the values that are repeated and using the time stamp of the first time that value was reported. After all, we know where the plane ended, and the points selected draw a line there over the time elapsed.
But if you want to do fine -grained analysis, the data won't work. Something about the laws of physics.