Automatic Cryptic Crossword Solver
John-Paul Knight
the solving of a cryptic crosword on a computer, using the
clues rather than an exclusive use of pattern matching
on checked sqaures. I am not aware that such a program
has been attemted before. However I believe it should
be possible from the fact that surface meaning in such
clues does not often have a direct impact on the solution,
i.e. there are not too many &lit. clues.
The program performs a simple lexical analysis
and parsing using keywords such as:
Anagram indicators,
hidden word indicators,
Container and contents indicators....
and parsing using the form...
Definition linkword Cryptic-Construction
or vice-versa
where linkwords are used to aid clue categorisation.
There is quite a lot of pattern matching used in later
stages of the program when simpler clues (i.e anagrams)
have been solved, this, I think, simulates the human solver.
I am currently in the specification stage and wondering how
far I should go with aspects such as I/P - O/P e.g a GUI
This is only a very brief sketch of the project.
I would like to hear any comments about this project and any
suggestions/ideas as to strategies in clue analysis. I would
also like to know if there is any information online that
may contain comprehensive lists of
a) common crossword jargon - e.g. Sailor = jack, abs , rn . etc.
b) common (and rare) clue indicators-e.g. Anagrams (upset, doctor)
hidden word (some, hides) etc.
If any one is interested, I will post regularly to this newsgroup
about my progress and upon completion (Easter 96) I will make the
program available as source so you can all see what kind of mess
I get in :-D
J.P Knight
Sik Cambon Jensen
constructs crossword puzzles. Though the problem is quite
different I guess we do have something in common. You find
solutions to cryptic clues I produce cryptic clues to solutions.
My problem is *easier* than yours since I don't have to find
clues that fit into squares etc. Still I need to be able to
adjust my cryptic clues making the overall puzzle 'the same
level'. If I want to give my puzzle a theme I too need a program
that can adjust clues. I'm not sure but a guess that some of
your solutions might solve mine problems and visa versa.
>I am currently designing and writing a program to simulate
>the solving of a cryptic crosword on a computer, using the
>clues rather than an exclusive use of pattern matching
>on checked sqaures. I am not aware that such a program
> has been attemted before.
My superviser on this project have a friend who have done
something close to that (in danish).
>However I believe it should
>be possible from the fact that surface meaning in such
>clues does not often have a direct impact on the solution,
>i.e. there are not too many &lit. clues.
You are probably right about single-word clues, but aren't
cryptic clues formed by sentences hard to solve?
>The program performs a simple lexical analysis
>and parsing using keywords such as:
>
>Anagram indicators,
>hidden word indicators,
>Container and contents indicators....
>
>and parsing using the form...
>
>Definition linkword Cryptic-Construction
>or vice-versa
>
>where linkwords are used to aid clue categorisation.
>
>There is quite a lot of pattern matching used in later
>stages of the program when simpler clues (i.e anagrams)
>have been solved, this, I think, simulates the human solver.
>
Interesting. Is it possible for me to get a brief description
of how you do these things?
>I am currently in the specification stage and wondering how
>far I should go with aspects such as I/P - O/P e.g a GUI
I guess it's hard to draw the line here. To begin with you
want to have the most sofisticated interface and at the end
you end up with a bare i/o interface with only the absolute
neccesary (I'm speaking for my self). If you have access to
Tcl and Tk that could make life easier. I plan to use it.
>This is only a very brief sketch of the project.
>
>I would like to hear any comments about this project and any
>suggestions/ideas as to strategies in clue analysis.
Try some of the papers below
1) XENO: Computer-Assisted Compilation of Crossword Puzzles, A
P.D. Smith, The Computer Journal, 26/4, pp 296-302, 1983
2) Computer Assisted Analysis of Cryptic Crosswords, P.W. Williams,
D. Woodhead, The Computer Journal, 22/? pp 67, 1979
3) Automation of Crossword Puzzle Solutions, G. Harris, J. Spring, J
Division Of CAD, Griffith University, Nathan, Australia, Manuscript?,
1979
4) Automatic Crossword Compilation and Solution, E.J. Jushvol, B.Sc
Thesis. Herriot-Watt University, Edinburgh, Scotland., 1984
5) Microprocessor Compilation and Solution of Crosswords, R.H. Davis,
E.J. Jushwol, Microproc. Microsystems, 9/10 pp 500-506, 1985
6) PICCUP: a Program for the Interpretation of Cryptic Crossword
Clues, A. M. Hart, M.Sc Thesis Herriot-Watt University, Edinburgh,
Scotland., 1991
7) Cryptic Crossword Clue Interpreter, A M. Hart, R.H. Davis,
Information and Software Technology, 34/1 pp 16-27, January 1992
I've seen 1) and 2) myself. If you see any of the others PLEASE let
me know.
>I would
>also like to know if there is any information online that
>may contain comprehensive lists of
>
>a) common crossword jargon - e.g. Sailor = jack, abs , rn . etc.
>b) common (and rare) clue indicators-e.g. Anagrams (upset, doctor)
> hidden word (some, hides) etc.
>
I know of one place where they have words/names/movietitles etc
available in 23 languages!
href="ftp://ftp.ox.ac.uk:/pub/wordlists/
>If any one is interested, I will post regularly to this newsgroup
>about my progress and upon completion (Easter 96) I will make the
>program available as source so you can all see what kind of mess
>I get in :-D
Well I'm interested.
SIK
-----------------------------------------------------------------------
email: s...@imada.ou.dk
www : http://www.imada.ou.dk/~sik
The CrossWord Puzzle Lab :
http://www.imada.ou.dk/~sik/Crossword_page/CrossMain.html
-----------------------------------------------------------------------
co...@ibl.bm
> I am currently designing and writing a program to simulate
> the solving of a cryptic crosword on a computer, using the
> clues rather than an exclusive use of pattern matching
> on checked sqaures. I am not aware that such a program
> clues does not often have a direct impact on the solution,
> i.e. there are not too many &lit. clues.
>
> and parsing using keywords such as:
>
> Anagram indicators,
> hidden word indicators,
> Container and contents indicators....
>
> and parsing using the form...
>
> Definition linkword Cryptic-Construction
> or vice-versa
>
> where linkwords are used to aid clue categorisation.
>
> There is quite a lot of pattern matching used in later
> stages of the program when simpler clues (i.e anagrams)
> have been solved, this, I think, simulates the human solver.
>
> far I should go with aspects such as I/P - O/P e.g a GUI
>
>
> I would like to hear any comments about this project and any
> also like to know if there is any information online that
> may contain comprehensive lists of
>
> a) common crossword jargon - e.g. Sailor = jack, abs , rn . etc.
> b) common (and rare) clue indicators-e.g. Anagrams (upset, doctor)
> hidden word (some, hides) etc.
>
> about my progress and upon completion (Easter 96) I will make the
> program available as source so you can all see what kind of mess
> I get in :-D
>
>
>
>
>>>>
Why?
co...@ibl.bm
<snip all that stuff about designing a program to solve crossword puzzles>
> >>
> >>
> >>>>>
> >Why?
> >
> >
> Why not?
>
>
Because the point of doing a crossword is to occupy one's brain for a while. To
have fun THINKING about it. Your program would negate all this. You might
as well just look at the answers and write them in.
coops
>>>>
David Squire
Ah yes coops, but that doesn't alter the fact that it's quite an interesting
problem from a sort of AI point of view. Given the current state of
knowledge with natural language parsing, I'd be pretty impressed and
interested in seeing a working solution.
Squizz
--
Squizz (David Squire)
Check out "http://www.cs.curtin.edu.au/~squizz/" for cv, research details,
humorous articles, cryptic crosswords and more....