How do I manipulate parse trees?

Question

I\'ve been playing around with natural language parse trees and manipulating them in various ways. I\'ve been using Stanford\'s Tregex and Tsurgeon tools but the code is a mess

Answer 1

Here is a second version in Common Lisp. This time I'm using a pattern matcher.

I'm using a function that matches a pattern against Lisp data. PMATCH:MATCH is an enhanced version of a pattern matcher found in the book Winston/Horn, Lisp, 3rd Edition. There are similar pattern matching functions available.

The data is as in my other answer.

The tree mapping function is changed to use the pattern matcher. The PMATCH:MATCH function returns T or an assoc list of bindings if the match is successful. It returns NIL if the match is not successful. The PMATCH:INSTANTIATE-PATTERN takes a pattern and a set of bindings. It returns a new list structure, where the pattern variables are replaced with the bindings.

(defun treemapp (tree pattern transformer)
  (cond ((null tree) nil)
        ((consp tree)
         (let ((bindings (pmatch:match pattern tree)))
           (if bindings
               (pmatch:instantiate-pattern transformer bindings)
             (cons (node-type tree)
                   (mapcar (lambda (child)
                             (treemapp child pattern transformer))
                           (node-children tree))))))
        (t tree)))

The example uses now patterns.

The pattern is a list structure. #?symbol matches a single item and creates a binding for symbol. #$symbol matches a list of items and creates a binding for symbol.

The transformer is a pattern that will be instantiated with the bindings.

(defun example1 ()
  (pprint (treemapp *tree*
                    '(NP (NP (#?type bank)) (PP #$children))
                    '(NP (NP (#?type bank) #$children)))))

Running this code returns the same result as in my other answer.