使用Python的urllib2和Beautifulsoup抓取维基百科时去除HTML标签

Question

我正在尝试抓取维基百科的数据，以便进行文本挖掘。我使用的是Python的urllib2和BeautifulSoup库。我的问题是：有没有简单的方法可以去掉我读取的文本中那些不必要的标签（比如链接的'a'标签或'span'标签）呢？

在这个情况下：

import urllib2
from BeautifulSoup import *
opener = urllib2.build_opener()
opener.addheaders = [('User-agent', 'Mozilla/5.0')]
infile = opener.open("http://en.wikipedia.org/w/index.php?title=data_mining&printable=yes")pool = BeautifulSoup(infile.read())
res=pool.findAll('div',attrs={'class' : 'mw-content-ltr'}) # to get to content directly
paragrapgs=res[0].findAll("p") #get all paragraphs

我得到的段落中有很多引用标签，比如：

paragrapgs[0] =

<p><b>Data mining</b> (the analysis step of the <b>knowledge discovery in databases</b> process,<sup id="cite_ref-Fayyad_0-0" class="reference"><a href="#cite_note-Fayyad-0"><span>[</span>1<span>]</span></a></sup> or KDD), a relatively young and interdisciplinary field of <a href="/wiki/Computer_science" title="Computer science">computer science</a><sup id="cite_ref-acm_1-0" class="reference"><a href="#cite_note-acm-1"><span>[</span>2<span>]</span></a></sup><sup id="cite_ref-brittanica_2-0" class="reference"><a href="#cite_note-brittanica-2"><span>[</span>3<span>]</span></a></sup> is the process of discovering new patterns from large <a href="/wiki/Data_set" title="Data set">data sets</a> involving methods at the intersection of <a href="/wiki/Artificial_intelligence" title="Artificial intelligence">artificial intelligence</a>, <a href="/wiki/Machine_learning" title="Machine learning">machine learning</a>, <a href="/wiki/Statistics" title="Statistics">statistics</a> and <a href="/wiki/Database_system" title="Database system">database systems</a>.<sup id="cite_ref-acm_1-1" class="reference"><a href="#cite_note-acm-1"><span>[</span>2<span>]</span></a></sup> The goal of data mining is to extract knowledge from a data set in a human-understandable structure<sup id="cite_ref-acm_1-2" class="reference"><a href="#cite_note-acm-1"><span>[</span>2<span>]</span></a></sup> and involves database and <a href="/wiki/Data_management" title="Data management">data management</a>, <a href="/wiki/Data_Pre-processing" title="Data Pre-processing">data preprocessing</a>, <a href="/wiki/Statistical_model" title="Statistical model">model</a> and <a href="/wiki/Statistical_inference" title="Statistical inference">inference</a> considerations, interestingness metrics, <a href="/wiki/Computational_complexity_theory" title="Computational complexity theory">complexity</a> considerations, post-processing of found structure, <a href="/wiki/Data_visualization" title="Data visualization">visualization</a> and <a href="/wiki/Online_algorithm" title="Online algorithm">online updating</a>.<sup id="cite_ref-acm_1-3" class="reference"><a href="#cite_note-acm-1"><span>[</span>2<span>]</span></a></sup></p>

有没有什么好主意可以去掉这些标签，只保留纯文本呢？

Answer 1

这些代码似乎是针对Beautiful Soup中的标签节点进行操作的。父节点会被修改，这样相关的标签就会被移除。同时，找到的标签也会以列表的形式返回给调用者。

@staticmethod
def seperateCommentTags(parentNode):
    commentTags = []
    for descendant in parentNode.descendants:
        if isinstance(descendant, element.Comment):
            commentTags.append(descendant)
    for commentTag in commentTags:
        commentTag.extract()
    return commentTags

@staticmethod
def seperateScriptTags(parentNode):
    scripttags = parentNode.find_all('script')
    scripts = []
    for scripttag in scripttags:
        script = scripttag.extract()
        if script is not None:
            scripts.append(script)
    return scripts

Answer 2

---

for p in paragraphs(text=True):
    print p

另外，你可以用 api.php 来替代 index.php：

#!/usr/bin/env python
import sys
import time
import urllib, urllib2
import xml.etree.cElementTree as etree

# prepare request
maxattempts = 5 # how many times to try the request before giving up
maxlag = 5 # seconds http://www.mediawiki.org/wiki/Manual:Maxlag_parameter
params = dict(action="query", format="xml", maxlag=maxlag,
              prop="revisions", rvprop="content", rvsection=0,
              titles="data_mining")
request = urllib2.Request(
    "http://en.wikipedia.org/w/api.php?" + urllib.urlencode(params), 
    headers={"User-Agent": "WikiDownloader/1.2",
             "Referer": "http://stackoverflow.com/q/8044814"})
# make request
for _ in range(maxattempts):
    response = urllib2.urlopen(request)
    if response.headers.get('MediaWiki-API-Error') == 'maxlag':
        t = response.headers.get('Retry-After', 5)
        print "retrying in %s seconds" % (t,)
        time.sleep(float(t))
    else:
        break # ready to read
else: # exhausted all attempts
    sys.exit(1)

# download & parse xml 
tree = etree.parse(response)

# find rev data 
rev_data = tree.findtext('.//rev')
if not rev_data:
    print 'MediaWiki-API-Error:', response.headers.get('MediaWiki-API-Error')
    tree.write(sys.stdout)
    print
    sys.exit(1)

print(rev_data)

输出

{{Distinguish|analytics|information extraction|data analysis}}

'''Data mining''' (the analysis step of the '''knowledge discovery in databases..

Answer 3

这是你可以使用lxml（还有很棒的requests库）来实现的方法：

import requests
import lxml.html as lh
from BeautifulSoup import UnicodeDammit

URL = "http://en.wikipedia.org/w/index.php?title=data_mining&printable=yes"
HEADERS = {'User-agent': 'Mozilla/5.0'}

def lhget(*args, **kwargs):
    r = requests.get(*args, **kwargs)
    html = UnicodeDammit(r.content).unicode
    tree = lh.fromstring(html)
    return tree

def remove(el):
    el.getparent().remove(el)

tree = lhget(URL, headers=HEADERS)

el = tree.xpath("//div[@class='mw-content-ltr']/p")[0]

for ref in el.xpath("//sup[@class='reference']"):
    remove(ref)

print lh.tostring(el, pretty_print=True)

print el.text_content()