Number Five (is alive)

On Monday, a crossword puzzle I created was published in the New York Times -- number five for the Times (2 Mondays, 1 Tuesday, a Thursday, and a Sunday) and number 35 or so overall. This one was not nearly as clever as my last one in the Times, but it was a Monday puzzle so it was accessible to a wider audience. (For a review of the puzzle click here and go to the February 2nd puzzle.) I like the idea of making a big name for myself in the world of crossword puzzle constructing (especially with NY Times editor Will Shortz), but I still have a way to go yet. Still, all I need to make it happen is a few more hours of free time each week and a bunch of kick-ass theme ideas. No sweat.

I use the software Crossword Puzzle Compiler for making my puzzles. Sometimes for fun I'll will have the computer auto-fill certain sections. I never actually use what the computer suggests though, for two reasons. One, I would feel somewhat fraudulent if I attached my name to a puzzle that a computer had a prominent part in constructing, and two, I can always do better than the computer. Sometimes it will tell me that the section cannot be filled and other times it fills it, but only after using a slew of obscure words. Things like IXIA and ELATER and SNED. Words that can be great to use in Scrabble, but are terrible to use in crossword puzzles. The reason for this, I imagine, is twofold: the word list that CPC uses is not very extensive -- it probably doesn't have many proper names (e.g., AESOP) nor many multi-word phrases (e.g., SUREENOUGH), and the algorithm CPC uses probably does not make much of a distinction between good crossword puzzle words (e.g., JAZZMEN, SHOWBIZ) and bad ones (e.g., IXIA, ELATER, SNED).

All of this got me thinking, what if you made a giant database of words and phrases and then gave everything on the list a score -- good crossword puzzle fill would get a high score, so-so fill a medium score, and bad fill a low score -- then, you wrote a program to fill a grid in a way that maximizes the sum of the scores of all the entries. Surely then a computer could make a very good crossword puzzle, at least as good, maybe even better than a human. When I thought of this I was very excited, thinking I had come up with a great original idea, but then I quickly thought, "wait, somebody has probably already had this idea." I did a Google search, and sure enough, I found this article.

Even though it meant my idea was not original I still found this article quite interesting. Also, I think crossword puzzle constructing will stay primarily a human endeavor for a long time to come, because most crossword puzzles rely heavily on a theme, and I don't think computers will be coming up with good themes any time soon. The traits computers don't have, creativity, cleverness, wit, humor are exactly the traits one needs to develop a decent theme. Sure, one could use a computer to fill a grid after the theme words have been entered, but c'mon, where is the fun in that?