An unconstrained crossword puzzle is one for which only the puzzle grid dimensions and the relevant dictionary are known. An algorithm to provide solutions for such an unconstrained crossword puzzle must, therefore, determine not only the locations of words in the word slots, as in the constrained problem, but also the positioning of the word slots themselves and, by implication, the location of the black squares. The implementation of such an algorithm is presented and demonstrated to be successful.
Harris, G. (1990). Generation of solution sets for unconstrained crossword puzzles. In Proceedings of the 1990 Symposium on Applied Computing (pp. 214-219). IEEE Computer Society. https://doi.org/10.1109/SOAC.1990.82171