This article should not be considered here as a "construction article", but can be used to give you a better understanding of how the antenna is designed. The final actual "construction" of the antenna will have to be left up to your ingenuity. 
referring to the original article if you can find it will be very helpful!

 As you can see in the drawing above, the overall length from end to end is about 16 feet or half of the normal length of a 20 meter dipole. The drawing is meant to be an "electrical" drawing rather than a "physical" one.The "active" elements are colored in orange including the main winding in the center. The black coupling winding, not shown with all it's windings, is centered over the main winding and is not offset as shown in the drawing.
The length from the outer end of the main winding to the outer edge of the 2 "arms", is equal to about 1/8 wavelength at 20 meters. So what we have here is 2 main conductive "arms" on each side of the main winding, connected with wire from top to bottom making a sort of triangular delta loop on each half of the antenna with the winding (coil) in the center and then the coupling loop is wound centered over that main winding.
The junction of each half of the antenna at the center of the antenna is soldered to the end of the main winding and one side of the antenna is used to connect the swr tuning tap point. An ohm meter would show a direct "short" from one side of the antenna to the other half....this is normal with this type of design due to the coil being in the center in series with each half of the antenna!
 
A workable "formula" can be derived for the element lengths as follows:
The original design center frequency was 14.15mhz in the 20 meter band.
1/8 wavelength = 234 / 2 = 117

117 / 14.15mhz = 8.26 feet per arm. (14.15mhz was the original design frequency in 1957 article in QST) This will be very close for experimental purposes and any small difference can be adjusted using the tap point.

The article and resulting research did not reveal the "angle" of the arms relative to any reference point but did say that the arms are "spread" so that the ends would be 5 feet apart. This does not exactly agree with the length of the connecting wire in the drawing above, between the arm ends, so it would be assumed that the wire length connecting the arms would be at least 5 feet! It is also assumed that the angle would not be extremely critical. The main elements were made from aluminum tubing in the original article published in QST. 3/4 inch aluminum tubing was used in the original.

Coil winding and tap point details.