Codeco parser in DCG

Status

Works but tokenization should be improved:

• split off punctuation
• don't be affected by trailing whitespace (the current work-around is to pipe the input through `sed "s/ *$//"`)

Files

• grammar.pl: parser in Codeco format (incl. a small test vocabulary), comes from the AceWiki distribution, but modified
• grammar_dcg.pl: parser in DCG (automatically generated, do not edit!)
• compile.sh: converts grammar.pl into grammar_dcg.pl, only needed if you modify grammar.pl
• parse.pl: parses STDIN using grammar_dcg.pl
• sentences.txt: example input for parse.pl

Modifications to grammar.pl

1. Add support for:
   • `isn't`
   • `aren't`
   • `don't`
   • `doesn't`
2. Every of-noun is also a noun and vice versa. This reflects better ACE where there is no explicit of-noun category.
3. Add support for `nothing but`, because this construct is used in the Ontograph testset.
4. `it_is_false that` now takes simple_sentence_1 as its argument (was: simple_sentence_2), i.e. `it is false that there is ...` is now possible, as well as `it is false that it is false that ...`.

Usage

The output (STDOUT) shows for each input sentence on a separate line its tokenization (where two spaces mark the token border) and a prefix OK or FAIL depending on whether parsing succeeded or not.

echo "Mary is a friend of Mary ." | sh run.sh

OK: Mary  is  a  friend of  Mary  .

cat sentences | sh run.sh

OK: Mary  asks  Mary  .
OK: Mary  is  a  friend of  at least  2  women  .  Mary  asks  Mary  .
FAIL: this  sentence  should  fail  to  parse
OK: everybody  asks  somebody  .
...

Parsing the complete test set (19k sentences):

time cat sentences | sh run.sh > out.txt

real	0m21.224s
user	0m21.097s
sys	0m0.144s