编辑：在我尝试了第一个答案之后，我仍然会在中断标签时得到文本中断（就像以前一样）。代码如下：

for event, element in ET.iterparse(path):
        if element.tag == "idsText":
            # move sentences from stack
            if sents:
                cat_texts.extend(sents)
                visited[current_doc] = visited.get(current_doc, 0) + 1
                # another function call
                annotate(cat_texts, filename, current_doc, visited[current_doc])
                # reset cat_texts
                cat_texts = []
            # set new current document's name
            current_doc = element.get("n")
            sents = []
        # new sentence starts
        elif element.tag == "s" and type(element.text) == str:
            if element.text.strip():
                sentence = ' '.join(element.itertext())
                new_sent.extend(nltk.word_tokenize(sentence, language='german'))
                sents.append(new_sent)
                new_sent = []

        element.clear()

我有一些非常大的xml文件（> 6 GB），想从中提取文本。问题是文本被其他标记打断。这里有一个虚拟文件，让你有一个想法：

<doc>
    <s> Here you can see a <ref target="SOME_URL" targOrder="u">reference</ref> in the text. </s>
</doc>

我想得到这个：

Here you can see a reference in the text.

我使用以下代码（尽管我在这里省略了细节）：

import xml.etree.ElementTree as ET
for event, element in ET.iterparse(path):
    if element.tag == "doc":
        # do something here            
    
    elif element.tag == "s" and type(element.text) == str:
        if element.text.strip():
            # again do something here

    element.clear()

应用于虚拟文件的代码将产生以下结果：

Here you can see a

我知道itertext（）会产生我想要的输出：

import xml.etree.ElementTree as ET
myxml = '<doc><s> Here you can see a <ref target="SOME_URL" targOrder="u">reference</ref> in the text. </s></doc>'
tree = ET.fromstring(myxml)
print(''.join(tree.itertext()))

输出：

Here you can see a reference in the text.

但是我想不出一种方法可以将它与iterparse（）（或任何其他增量解析方法）结合起来。因为我不能一次将xml解析成树，因为它的大小。增量地执行它意味着itertext（）将不起作用，因为当解析带有标记的元素时，下面的标记（在本例中为<ref）还没有被解析。
有没有一种方法可以获取元素中的所有文本，并在增量解析时剥离标签？
非常感谢！