>Ran across this today and thought it was interesting.
There are fairly old speech vertical and horizontal human speech polars by Dunn and Farnsworth which the article does not seem to mention. The vertical ones are quite *funky*.
If you get hold of the polars, you could make your own EASE model. If would be cool if it was included, but i guess the origin and use of EASE is mostly for loudspeaker electro-acoustical modelling, not acoustics as such.
The sheer resolution of EASE v3/v4 1/3rd octave 5 degree resolution makes making a speaker fron scratch a daunting task.
I would do it in the old ease v2 (1 octave, 10 degree reso). If you do not have it, you can do it on EASE itself, maybe help yourself with the spreadsheet on the service area. You will not find oblique angle data, just horizontal and vertical, so wou'll have to use EASE's eliptical interpolation.