Try this:

Write a procedure that tries to match each of the words in the list, only at the very beginning of the string. If there is a match, the procedure calls itself to match the remainder of the string, otherwise it just returns. Except... if the remainder of the string is zero length, print the segments that matched.

The parameters of the procedure could be the sequence of words that have have been matched so far separated by dashes, and the remaining part of the string.

For your example, you would start with

try_match( "", "cowboycatdog" )
[download]

which would, in one of the intermediate steps, call itself with the parameters

try_match( "cow-boy", "catdog" );
[download]

and later,

try_match( "cow-boy-cat-dog", "" );
[download]

Assuming the number of different words in the list is about proportional to the string length, the run time should be of the order of the cube of the length of the string, I think.