python - 树字符串到括号字符串

Question

假设我有一个句子的树字符串:

s = "(TOP (S (NP-TMP (NP (DT This) (NN time)) (ADVP (RP around))) (NP-SBJ (PRP they)) (VP (VBP 're) (VP (VBG moving) (ADVP (RB even) (RBR faster))))))"

我想将它转换成这样的括号结构:

"(((This time)(around))(they)(('re)((moving)(even faster))))"

我尝试执行以下操作:

import nltk

s = "(TOP (S (NP-TMP (NP (DT This) (NN time)) (ADVP (RP around))) (NP-SBJ (PRP they)) (VP (VBP 're) (VP (VBG moving) (ADVP (RB even) (RBR faster))))))"
tree = nltk.Tree.fromstring(s)

out = "("
for subtrees in tree:
    # there are threee subtrees
    # print(len(subtree))
    for i, subtree in enumerate(subtrees):
        if len(subtree) > 1:
            out += "("
        for bracketing in range(len(subtree)):
            # print(subtree[bracketing])
            flattened_tree = subtree[bracketing].flatten()
            flattened_string = str(flattened_tree)
            flattened_string = flattened_string.replace(flattened_tree.label() + " ", "")
            print(flattened_string)
            out += flattened_string
        if len(subtree) > 1:
            out += ")"
        # break
out += ")"

print(out)
# (((This time)(around))(they)(('re)(moving even faster)))

编辑:

如果您看到的话，“This” 和 “time” 是同一个父项 “NP” 的一部分。因此，它们成为连续的组成部分，即(这次)。

然而，“around” 是单个单词组成部分，尽管是同一左子树的一部分。所以，它变成了((这次)(大约))。

类似地，对于右子树的情况 - “'re” 和 “'移动得更快”，我们看到 “移动” 以及“甚至更快” 共享同一个父级，“VP”。

所以，它变成了，(('re)((移动)(甚至更快))。

Answer 1

如果我理解正确的话，当原子字符串有一个不是原子字符串但已经是括号组合的同级时，您希望向原子字符串添加括号，这样您就永远不会有这样的模式:

(moving (even faster))

或者换句话说，空格只能作为两个原子字符串之间的分隔符。

我会通过两次递归来完成此操作:

• 第一个将树转换为嵌套列表的方法。这将更容易区分上述规则。

• 第二次将该嵌套列表转换为最终的括号字符串。

代码:

import nltk

def tolist(tree):
    if isinstance(tree[0], str):
        return tree[0]
    res = [tolist(subtree) for subtree in tree]
    leaves = sum(isinstance(child, str) for child in res)
    if 0 < leaves < len(res): # if there is a mix of leaves and subtrees....
        # ...then wrap every leaf in a list
        return [[child] if isinstance(child, str) else child for child in res]
    return res

def tostr(lst):
    if isinstance(lst[0], str):  # assume all list members are strings
        return "(" + " ".join(lst) + ")"
    return "(" + "".join([tostr(sub) for sub in lst]) + ")"

# run on sample data
s = "(TOP (S (NP-TMP (NP (DT This) (NN time)) (ADVP (RP around))) (NP-SBJ (PRP they)) (VP (VBP 're) (VP (VBG moving) (ADVP (RB even) (RBR faster))))))"
tree = nltk.Tree.fromstring(s)
res = tostr(tolist(tree[0]))
print(res)