
(Message 85)
From:     HELEN    HPS (on ERCC DEC-10) <Pain@EDXA>
Date:     Wednesday, 30-Jan-85 22:59:54-GMT
To:       ecmi04@2972,Pain@EDXA
Via:      uk.ac.edinburgh.edxa  ; (to uk.ac.edinburgh.emas) 30 Jan 85  23:01:57 gmt
Msg ID:   <132017-752-505@EDXA>

--------


            AI1 PROLOG PROJECTS - NATURAL LANGUAGE PROJECT            AI1 PROLOG PROJECTS - NATURAL LANGUAGE PROJECT            AI1 PROLOG PROJECTS - NATURAL LANGUAGE PROJECT


                                                            Helen Pain
                                                       30 January 1985


  This  project is based on a program that generates sentences from an
recursive transition network (RTN).  The  RTN  representation  of  the
grammar  consists of a number of states and connecting arcs.  The arcs
are labelled by other RTN's (non-terminals) or by  words  (terminals).
The  RTN  may  be traversed and a sentence generated in the process of
traversal.  To run the program three files need to be consulted:  

                'ecmi04.rtn'
                'ecmi04.drinks'
                'ecmi04.random'

  The file 'ecmi04.rtn' contains the sentence  generating  predicates,
generate and traverse.

  The  file 'ecmi04.drinks' contains the grammar itself - it comprises
RTN's for sentence, nounphrase,  verb,  noun,  determiner,  adjective,
proper_noun and stop_mark.

  The file 'ecmi04.random' contains number of predicates that are used
to select an element from a list, at random. (NOTE: you do not have to
worry  about  how  the predicates in this file work - you only need to
know that the predicate random_pick(List,El) matches List to a list of
any number of elements  and  that  El  matches  to  a  random  element
selected from that list).

  To generate a sentence, in response to the prompt type 

	?- go.

Try  this several times and explore the sentences generated.  List the
program. Trace the program and step through it.  

	?- debug go.

When you think that you understand how  the  program  works  do  the
following:
   1. Write  a description of how the program works:  a paragraph
      for each of the procedures in the program, describing  what
      each procedure does, and how it does it.

   2. Draw  the  RTN's  that  are  represented  by  the  'drinks'
      grammar.  [see your linguistic notes].

   3. Provide 6  examples  of  the  sentences  generated  and  an
      example trace of the program for two of these examples.                                  2

   4. Copy  the  "drinks"  file to your area.  Edit it to include
      the following additional vocabulary:
                verbs - hates,pours
                nouns - pint, milk, glass
                adjectives - clear, fresh, full
      Include a  listing  of  the  editted  grammar,  and  sample
      sentences produced, in your project report.

In  addition  to  the  above,  do  at  least  two  of  the following
extensions to the project. For each, include in your project report:
   i. a description of the extension to  the  program  (including
      any new RTN's used);
  ii. a listing of the program to indicate the changes made;
 iii. a number of examples to illustrate the extension;
  iv. any  relevant discussion of the problem; how you solved it,
      how else you might have solved it, where it fails and  why,
      etc.

_______ __________Program_Extensions

1.  This program does not distinguish between subject and object. It
can produce sentences such as:  
       "the tea drinks the vicar"
Rewrite the RTN's to prevent this.  Rewrite the  program  (edit  the
"drinks" file) to include these new rtn's.

2. How could the program be altered to generate the following: 
      fred like a pint of beer.
      the vicar pours a glass of sherry!
Add  to the sentence grammar an RTN for prepositional phrases.  Change
the program to generate sentences  that  include  these  prepositional
phrases. Try also adding conjunctions, for example:  
      jane wants a vodka and tonic.
      the policeman drinks the tea and the beer?
Alter the program to generate these.

3.  Copy  the 'rtn' file and edit it so that a call of "go" prints a
parse tree in the form:
          sentence
            nounphrase
                determiner
                    a
                noun
                    vicar
            verb
                drinks
            nounphrase
                determiner
                    the
                noun
                   tea
            stop_mark                                  3

[harder extensions]

4. Rewrite the drinks grammar and rtn to produce poems. (They do not
have to rhyme!)

5. Rewrite the program so that, instead of generating a sentence, it
takes a sentence, for  example  [the,vicar,loves,tea],  and prints a
parse  of it  using the grammar - or fails if it will not parse. [it
might produce the same sort of parse tree as that above].

6.  Find  out  about  the  Prolog  notation  for  "Definite   Clause
Grammars".  Rewrite the sentence generator using these. 

Please  provide  TWO  copies  of  your project report, including all
listings and examples.  These should be submitted to me, Helen Pain,
by Monday 22nd April (that is, Monday of week 2 of next term).

--------
    