Change the value of the POS tag in NLTK Tree leaf

Question

I have an nltk.tree.Tree object.

t = Tree('S', [Tree('NP', [('I','tag')]), Tree('VP', [Tree('V', [('saw','tag')]), Tree('NP', [('him','tag')])])])

I want to traverse it with the function below, and change every leaf's POS tag (ie 'tag' in the above example).

def traverse(tree):
    try:
        tree.label()
    except AttributeError:
        tree[-1] = ('another_tag')
        print(tree)
    else:
        for child in tree:
            traverse(child)

Unfortunately, every POS-tag in the leaf cannot be changed because the tuple object including it is immutable.

How can I change the POS-tags in the example without affecting it's original tree structure?

I'm fairly new to this tree structure, please show some clear excerpts how to deal with nested structure.

Answer 1

An nltk tree is actually just a list. With enumerate, you can loop through it and assign the node at position ia new value. Something like:

def traverse(tree):

    for index, subtree in enumerate(tree):
        if type(subtree) == nltk.tree.Tree:
            traverse(subtree)
        elif type(subtree) == tuple:
            newVal = (subtree[0], subtree[1].lower())
            subtree = newVal
            tree[index] = subtree

Because you're dealing with tuples (immutable), you cannot replace only the POStag, but have to create a new tuple. The code above just makes the tag lowercase, but you can put in anything you like as second element of the newVal tuple.

Hope this helps!