October 2004. A brand new release of the clue solver technology is now live!
The author of the technology, William Tunstall-Pedoe, has spent several months implementing a number of major new backend features and many smaller improvements. Further improvements
are planned over the next few months.
The big improvements are:
1. Wild translations
In all previous versions of the technology, the software needed to know how to correctly interpret
every word of the word-play part of the clue in order to produce a coherent explanation of how it worked
(with the exception of some straight anagram and hidden word clues where it didn't need to recognise the
The latest version can now usually generate the correct answer when one translation is completely
(or partially) unknown. This means it can solve more cryptic clues and provide better explanations
than it could previously!
For example, for the clue:
Inform charmingly old-fashioned Bill first (8)
The correct answer is ACQUAINT but the translation "charmingly old-fashioned" -> "quaint" is not in
the software's knowledge base.
The old version of the technology would list the correct answer second (confidence 5%) with the explanation:
'inform charmingly' is the definition.
I am not sure about the 'charmingly' bit but 'acquaint' can be an answer for 'inform' (acquainting is a kind of informing).
'old-fashioned bill first' is the subsidiary indication.
I cannot quite see how this works, but
'bill' could be 'ac' ('bill'->'ac' is in my crossword knowledge base) and 'ac' is present in the answer.
'first' could be 'i' ('first'->'i' is in my crossword knowledge base) and 'i' is present in the answer.
The remaining letters 'quant' is a valid word which might be clued in a way I don't understand.
This may be the basis of clue (or it may be nonsense).
i.e. Although it has got the answer from the definition and can see the "bill"->"ac" bit, it's explanation
is rather poor. The new version however, gives the answer first with 98% confidence and the following explanation:
'inform' is the definition.
(acquainting is a kind of informing)
'charmingly old-fashioned bill first' is the subsidiary indication.
'charmingly old-fashioned' becomes 'quaint' (I can't justify this - if you can you should give a lot more credence to this answer.)
'bill' becomes 'ac' ('bill'->'ac' is in my crossword knowledge base).
'first' means one lot of letters go next to another.
'quaint' after 'ac' is 'acquaint'.
i.e. The explanation is as perfect as it can be considering the missing knowledge!
Another example is:
Basically, a nest egg meant for use by a friend (13)
The translation "a nest egg" -> "fund" is a little tenuous and not in the software's knowledge base
but this is no barrier to the new technology solving the clue with 100% confidence!
'basically' is the definition.
('fundamentally' can be a synonym of 'basically')
'a nest egg meant for use by a friend' is the subsidiary indication.
'a nest egg' becomes 'fund' (I can't justify this - if you can you should give a lot more credence to this answer.).
'for use' indicates an anagram (using the letters).
'by' means one lot of letters go next to another.
'a friend' becomes 'ally' ('ally' can be a synonym of 'friend').
'meant' anagrammed gives 'ament'.
When the missing knowledge seems plausible it can even adapt its explanation to say so and give
extra credence to the clue. For example:
Stop a retired cleric going to Thailand (5)
The correct answer is avert. The old technology listed this answer well down the list
of possible answers with no explanation other the definition. The new technology lists the correct answer
first with 97% confidence and can produce an almost perfect explanation with exactly the same knowledge
available to the old!
'stop' is the definition.
('stop'->'avert' is in my internal thesaurus)
'a retired cleric going to thailand' is the subsidiary indication.
'retired' shows that the letters should be reversed in order.
'cleric' becomes 'rev' (abbreviation for Reverend).
'going to' means one lot of letters go next to another.
'thailand' becomes 't' (this might be a standard abbreviation of which I'm unaware).
'rev' reversed gives 'ver'.
The new technology can see that the clue works perfectly if the translation "Thailand" -> "T" exists
and can also see that this might be an abbreviation! The fact it isn't in the knowledge base is no
barrier to it producing the right answer with a perfect explanation!
Similarly, the new technology can see that a translation may be true if it wasn't for the odd
adjective or extra words of which it is unaware!
Yet another advantage of this new approach is that it can sometimes overcome the "combinatorial explosion"
problem that sometimes affected the old version interpreting the wordplay correctly. If a word or phrase
has many translations and is part of a complicated subsidiary indication, the old version could sometimes
miss the correct interpretation just because there were too many to try. Although this can still
happen with the new version, it happens less frequently because the wild translation can be matched
with a correct translation later.
2. Definitional Hints
A whole new class of knowledge has been added with this release: definitional hints. These
are words that may appear in a non-cryptic clue or the definition part of a cryptic clue but which don't
by themselves define the answer. These could be adjectives or adverbs which relate to the answer or
simply nouns that are associated with the answer.
These address a draw-back with the old technology which was that a defininition needed to be defined directly
to get a direct answer. Often there are numerous ways of definining a word so unless every possibility was
listed, the answer would either not be produced or the explanation would have to ignore certain words.
The new source of knowledge
means that additional words can be used to reinforce an answer, give a better explanation and provide more confidence. e.g.
Spanish drink with fermentation of grain in it (7)
Now gives this explanation for its first choice, 100% answer, Sangria
'spanish drink' is the definition.
'sangria' can be an answer for 'drink' (sangria is a kind of drink). I also know that 'spanish' relates to this answer.
'fermentation of grain in it' is the subsidiary indication.
'fermentation of' indicates an anagram.
'in' means one lot of letters goes inside another.
'it' becomes 'sa' (abbreviation for Sex Appeal - crossword cliché).
'grain' anagrammed gives 'angri'.
'angri' placed inside 'sa' is 'sangria'.
'with' is the link.
Earlier versions of the technology would either need the direct translation "spanish drink" -> "Sangria" to be
in the knowledge base or would have to ignore the "spanish" bit altogether, get the answer from "drink" and
confess to not knowing the significance of the word "spanish".
Currently, there is very little definitional hint knowledge in the knowledge base so it is not yet
having a big effect on solving rates. However, we are adding new hint knowledge every day and are currently
looking at automatic ways to add large amounts of such knowledge automatically. You can also add
such knowledge yourself using the new Add Definitional Hint tool
Another feature of this release is that the clue "solving" tool can be used just to explain a cryptic
clue when you already know the answer. For example, when an answer is published in the newspaper the next
day and you don't understand how it works.
To use this feature you just enter the answer as a complete letter pattern (i.e. no '?' characters). The solver
will see you have done this and adapt its summary accordingly.
Judge taking tea-break after Times puzzle (8)
Letter pattern: estimate
and the clue solver will say
"I believe this is a perfect explanation of how the clue produces your answer."
and give the following explanation:
'judge' is the definition.
(estimating is a kind of judging)
'tea-break after times puzzle' is the subsidiary indication.
'break' indicates an anagram.
'after' means one lot of letters go next to another.
'puzzle' indicates an anagram.
'tea' anagrammed gives 'ate'.
'times' anagrammed gives 'estim'.
'ate' after 'estim' is 'estimate'.
'taking' is the link.
3. Improved letter pattern syntax
The syntax for entering letter patterns has been improved:
Spaces can be included
in the pattern.
Underscores "_" can be used in addition to question marks "?" to
indicate a single unknown letter.
Two digit numbers can now be used to indicate runs of unknown
letters longer than 9. (e.g. "a12e" is the same as "a????????????e").
4. Better handling of clues involving Removal indicators
A lot of work has gone into better handling of these. A higher percentage of such clues can now be solved.
In addition, the engine can now solve what we call "broken remover" clues. This is where the letters to be
removed (the "remover") are not present as a single sequence in the letters to remove from (the "removand"). Some
crossword editors allow broken remover clues, others (such as The Times)
disallow them. The solver can now solve both types.
Examples of removal clues that the new solver can solve include:
Having lost one Seles learnt to become persistent (10)
Unsettled comedian lacked the energy to be funny (7)
Swimmer in underwear abandoning the lake (4)
4. Other changes
Numerous other small changes have been implemented to almost everything the backend does. These result in
a higher percentage of clues that the solver can solve and the complexity of clues that can now be attempted.
The following is a very complicated cryptic clue that can now be solved (from a recent Private Eye crossword).
Note the shuddering domestic appliance Bill regularly installed, noisy thing (6,7)
Go to the Clue Solver now!