从 XML 标记创建原始文本

Question

我有一些通过 NLP 处理器运行的 XML。我必须在 Python 脚本中修改输出，所以对我来说没有 XSLT。我试图从我的 XML 中提取 <TXT> 和 </TXT> 中的所有原始文本作为字符串，但我一直坚持如何从 ElementTree 中提取它。

到目前为止我的代码是

import xml.etree.ElementTree as ET

xml_doc = """<?xml version="1.0" encoding="UTF-8"?>
<NORMDOC>
   <DOC>
      <DOCID>112233</DOCID>
      <TXT>
        <S sid="112233-SENT-001"><ENAMEX type="PERSON" id="PER-112233-001">George Washington</ENAMEX> and <ENAMEX type="PERSON" id="PER-112233-002">Thomas Jefferson</ENAMEX> were both founding fathers.</S>
        <S sid="112233-SENT-002"><ENAMEX type="PERSON" id="PER-112233-002">Thomas Jefferson</ENAMEX> has a social security number of <IDEX type="SSN" id="SSN-112233-075">222-22-2222</IDEX>.</S>
      </TXT>
   </DOC>
</NORMDOC>
"""

tree = ET.parse(xml_doc) # xml_doc is actually a file, but for reproducability it's the above xml

然后我想从那里提取 TXT 中的所有内容作为去除标签的字符串。它必须是后面一些其他进程的字符串。我想看起来像下面的output_txt。

output_txt = "George Washington and Thomas Jefferson were both founding fathers. Thomas Jefferson has a social security number of 222-22-2222."

我想这应该相当简单明了，但我就是想不通。我尝试使用this 解决方案，但我得到了AttributeError: 'ElementTree' object has no attribute 'itertext'，它会剥离xml 中的所有标签，而不是仅在<TXT> 和</TXT> 之间。

Answer 1

通常我会使用纯 XPath 来执行此操作：

normalize-space(//TXT)

但是，ElementTree 中的 XPath 支持是有限的，因此您只能在 lxml 中执行此操作。

要在 ElementTree 中执行此操作，我会执行类似于您在问题中链接到的答案；使用method="text" 将其强制为带有tostring 的纯文本。您还想规范化空白。

示例...

import xml.etree.ElementTree as ET

xml_doc = """<?xml version="1.0" encoding="UTF-8"?>
<NORMDOC>
   <DOC>
      <DOCID>112233</DOCID>
      <TXT>
        <S sid="112233-SENT-001"><ENAMEX type="PERSON" id="PER-112233-001">George Washington</ENAMEX> and <ENAMEX type="PERSON" id="PER-112233-002">Thomas Jefferson</ENAMEX> were both founding fathers.</S>
        <S sid="112233-SENT-002"><ENAMEX type="PERSON" id="PER-112233-002">Thomas Jefferson</ENAMEX> has a social security number of <IDEX type="SSN" id="SSN-112233-075">222-22-2222</IDEX>.</S>
      </TXT>
   </DOC>
</NORMDOC>
"""

tree = ET.fromstring(xml_doc)

txt = tree.find(".//TXT")
raw_text = ET.tostring(txt, encoding='utf8', method='text').decode()
normalized_text = " ".join(raw_text.split())
print(normalized_text)

打印输出...

George Washington and Thomas Jefferson were both founding fathers. Thomas Jefferson has a social security number of 222-22-2222.