Now online: codewordsolver.com

About a year ago I was stuck on a fiendish codeword. Not wanting to pay the extortionate rates that papers charge for puzzle helplines, I created a codeword solver.

During this process, I found it fascinating to study the way that words are constructed. English is unique in the vastness of its foreign influences, and several of the most difficult clues are loan words, like “verandah”. They do not follow the same letter-ordering logic that characterises most English vocabulary.

What goes through your mind when you see an incomplete word: r . p . . t . . ? Why does the full word suddenly jump into your head on some days, whilst on others nothing comes for hours?

The closest thing to codewords on television is the final round of the UK game show Only Connect. Cultural phenomena like book titles or famous sportspeople flash up with vowels deleted, i.e. “prd ndp rj dce” for Pride and Prejudice. Contestants score points by buzzing in first and calling out the full phrase. People tend to vary hugely in their performance in this round: half never buzz in, whilst others seem to be impossibly fast at seeing answers across all spheres of knowledge, like this guy.

I only get there quicker than the contestants when the category of the clues is something that I’m very familiar with. Otherwise, I’m way slower than them.

If I were to program a computer to be good at this game, it would have to contain a database of all culture. The program would then perform many operations to compare the missing vowels string against potential matches in the correct category. Although it doesn’t ‘feel’ like that’s what happening inside our brains when we complete codewords or find missing vowels, there must be a similar algorithm at work.

The computer program could be sped up a lot by indexing the potential matches by initial letter. Perhaps that’s why it’s so much harder to complete a clue when we lack the first letter.