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

Beauty is in the eye of the beholder. But you never say how the code of Tregex or Tsurgeon is a mess. It sounds more like you can't deal with Java or greater abstraction and so you're looking for something concrete written in Python.

There's nothing wrong with hand-writing tree matching and transformation functions. Indeed, we used to do that all the time. But after the first couple of hundred, it seemed like there had to be a better way, and hence we moved to using the domain-specific languages of Tregex and Tsurgeon. This is generally seen as a laudable programming style. See in Wikipedia. They're well-specified languages with an exact syntax specification, etc. Here is your example using them.

Tree t = Tree.valueOf("(ROOT (S (NP (NP (NNP Bank)) (PP (IN of) (NP (NNP America)))) (VP (VBD used) (S (VP (TO to) (VP (VB be) (VP (VBN called) (NP (NP (NNP Bank)) (PP (IN of) (NP (NNP Italy)))))))))))");
TregexPattern pat = TregexPattern.compile("NP <1 (NP << Bank) <2 PP=remove");
TsurgeonPattern surgery = Tsurgeon.parseOperation("excise remove remove");
Tsurgeon.processPattern(pat, surgery, t).pennPrint();

Note that the Java code is actually shorter than the Lisp code, precisely because of use of the domain-specific language. It's hard to see how this could be simpler: specify pattern, specify operation, apply.

But if you'd prefer to be hand-writing methods that match patterns on trees and change them into other trees in Python, then you're most welcome to go off and do that.