English plurals

A first attempt

Here is a start at building a system for pluralising English words. Run it by typing plural followed by some singular form of a noun, e.g. plural "cat. Don't forget the quotes.
to plural :noun
 if sibilant.end? :noun [op suffix :noun "es] [op suffix :noun "s]
 end
 
 to sibilant.end? :word
 if member? last :word [s z x] [op "true]
 if and ( last :word ) = "h not ( last bl :word ) = "t [op "true]
 op "false
 end
 
 to suffix :stem :affix
 op word :stem :affix
 end
 
Sibilant.end? does not really perform appropriately. It outputs "true if the input ends in s or z or h (as long as the h is not preceded by t). Some normal words therefore give it problems. Which? And can you edit sibilant.end? to deal with these cases correctly?

Plural as it stands does not handle correctly the plurals of words like daisy which end in y. Can you modify the procedure to take care of this problem? You will need to define a predicate consonant? and you might find it useful to take advantage of its companion called vowel?. You could perhaps also create a predicate ends.in.y?, but this would be less generally useful.

In the end, there are, of course, exceptions which are best treated by listing them. One well-known and amusing way to handle these is to set up for every such case a variable which has the singular form as its name and the plural form as its value. For example make "ox "oxen. With variables like this in place, we now edit plural so that it checks first to see if the word to be pluralised is a variable name and if so outputs the corresponding variable value without further ado:
to plural :noun
 if name? :noun  [op thing :noun]
 if sibilant.end? :noun [op suffix :noun "es] [op suffix :noun "s]
 end
The way the pluraliser now handles exceptions presupposes that all processing of words is from singular to plural. Assume that the system was required to convert not only singulars to plurals but also, on occasions, plurals to singulars. In these circumstances, how could the information about exceptions be more conveniently stored?

French dictionaries and articles

If you are unhappy with the fake version of entry, and would like it to output a range of words rather than a single word, one alternative is to rewrite the procedure using a series of conditional instructions each based on a test like :word = "fromage. Whenever a particular test returns "true, the dictionary entry of the corresponding word should be output. Notice tthat this tactic effectively means building the dictionary into the definition of entry rather than treating it as a separate data structure, as we originally proposed.

(Look here if you can't work out how to do this. But please experiment first!)

There are, of course, yet other ways to implement the dictionary in the first place. For example, each word could be set up independently (using make) as the name of a variable, the value of which is its complete dictionary entry.

E.g. make "fromage [[category noun] [gender masculine] [means cheese]]

Can you write a version of entry (assuming it still takes a word as input) which is adapted to this new arrangement?

How would you improve the treatment of French articles provided earlier in this section so that you could handle the fact that there are some French words like herbe 'grass' which behave as though they begin with vowels - l'herbe - while others (fewer) like hibou 'owl' behave as though they begin with consonants - le hibou?

Suppose - to take a different line again - that a bilingual dictionary is handled as a list of word pairs [[book livre][pen stylo] ... ]. Define a predicate correct.translation? which takes two inputs :word1 :word2 and makes use of member? to output "true when :word1 is the English equivalent of the French :word2 E.g. correct.translation? "book "livre should output "true and correct.translation? "book "stylo should output "false. What does your version of correct.translation? do with correct.translation "livre "book? If it bothers you, can you see how to correct the problem?



Ron Brasington
Department of Linguistic Science
The University of Reading
Reading
UK

E-mail: ron.brasington@rdg.ac.uk