The idea of this project is providing code to generate a certain special kind of crosswords sometimes called Literary Cryptograms. Rather than a complex description of how they look like, I think that an example will convey every relevant detail. Start by choosing a literary work and a quote from it, for instance:

"If we admit that human life can be ruled by reason, then all possibility of life is destroyed" Leo Tolstoy, War and Peace

The idea is to find a set of words such that:

• Their initials form the title of the work.
• Their letters can be so reordered as to form the quote.

A possible solution follows below:

Wishfulfilment
Audile
Ranty
Aridly
Nihili
Doost
Potboy
Estal
Asbestine
Chaffered
Embe,

where the letters should be permuted according to the 0-indexed numbers in brackets:
W (2) i (0) s (37) h (10) f (1) u (14) l (18) f (20) i (7) l (29) m (6) e (3) n (17) t (8);
A (4) r (27) i (19) d (5) l (45) y (33);
R (34) a (11) n (24) t (9) y (57);
A (16) u (28) d (31) i (51) l (46) e (21);
N (39) i (53) h (13) i (55) l (54) i (61);
D (66) o (38) o (48) s (49) t (12);
P (47) o (58) t (40) b (25) o (71) y (72);
E (26) s (50) t (56) a (23) l (60);
A (36) s (65) b (32) e (30) s (68) t (69) i (64) n (43) e (35);
C (22) h (41) a (44) f (59) f (62) e (42) r (70) e (63) d (74);
E (67) m (15) b (52) e (73).

For more details, check my post.

Warning: Rather than a mature and fully developed project, this repository is just intended as a proof of concept and you probably would need to write and modify code to suit your needs.

├── data                 <- folder holding English dictionary paired with values (according to its frequency) and vectorized codifications of words
├── optimization         <- folder holding integer programming model, input files and a Python script to run the solver
├── src                  <- folder holding some auxiliary Python scripts to generate the data. 

The dictionary of words paired with its relative frequencies was taken from this repository.

To generate a literary cryptogram, you will first need to fill the selected excerpt and title in optimization > excerpt.txt and optimization > title.txt, respectively. Once this is done, the following step depends on whether or not you have SCIP installed. SCIP is a fast non-commercial solver for mixed integer programming (MIP), but it requires a license for non academic purposes.

Simply go to the optimization folder and type > python solve_script.py excerpt.txt title.txt. In most cases, this would write the best solution found to optimization > cryptogram.txt. If this not be the case, you can try incrementing (or directly removing) the time limit imposed in solve_script.py.

Running as before > python solve_script.py excerpt.txt title.txt would obviously fail but it will leave prepared the integer program at optimization > model.zpl. Then, you can download Zimpl here and run > zimpl model.zpl. This will generate a model.lp file that can be fed and optimized by any other solver you have access to. Alternatively, running > zimpl -t mps model.zpl generates a model.mps file that can be read by PuLP.