I recently decided to take another look at my stalled Crossword Compiler word list project. I thought it would be worthwhile to analyze the strengths and weaknesses of the Autofill Project and modify my word list management to serve the specifications of my current primary puzzle construction outlet: Crosswords With Friends. Also, tinkering with word lists is always a good way to feel productive while procrastinating on other projects.
Autofill opponents generally cite lousy grid entries as the fundamental and inevitable problem of delegating construction duties to a computer. I was reluctant to jump on this bandwagon and believed that autofill could perform satisfactorily with a well designed word list. I groomed my list for years using a complex word-scoring protocol and eventually developed a data set that could reliably autofill small to medium grid sections. But among the lively fill I noticed subtle word duplications (ATECROW and EATERY), proper names with blind crossings, and grids with disproportionate sets of entries related to entertainment, science, or some other category. I couldn’t think of a way to fine-tune CCWIN to address all of these personal and industry-sanctioned aesthetic standards, so I finally accepted that grid fill requires human oversight.
In the new default word list I am confining entries to five scoring categories. A sophisticated scoring system with a 2-point differentiation between singular and plural nouns is not very important when I am manually reviewing all entries before they go in the grid. The simplified scoring system is making the conversion process easier, and I am already halfway through the scoring adjustments for six-letter entries. The other major change in the new default is reformatting entries to mixed case with spaces. Of the 366,000 entries in my all-caps default list, about 155,000 have been converted to mixed case. The unformatted entries are mostly long phrases and garbage that I don’t need for Crosswords With Friends, but I do want to convert the salvageable material from this list eventually.
While I am making these changes to my default list, I am not making efforts to add a lot of new entries. My current word list suits me fine. I add about a dozen entries a month from my notepad and I will gradually go through the lists that other constructors have sent me, but I don’t plan to search Google for theme lists or start mining AcrossLite files of published crosswords.