python - 当使用维基百科的 API 的 .random() 函数时，为什么我得到的页面越多，重复页面的数量就越多？

Question

我正在使用 Wikipedia 的 API 来抓取挪威语内容，对其进行清理并将其写入文件，以用于为 CMU Sphinx 训练语言模型。

在 for 循环中运行 .random-function 时，但我遇到了一个问题。我正在通过 pageId 计算唯一页面的数量，并且我得到了大量的重复。一开始并没有太多，但过了一段时间，重复的数量是唯一 ID 数量的两倍。当我们得到 40 页时，我们有大约 80 个副本。

当然，我们没有看到关于 .random-function 的某些内容？

这是代码。regEx 在函数中可以更轻松地读取过滤器顺序。

import re

import wikipedia

"""
Module wikitest.py - Script for scraping Wikipedia of text based on articles 
found by using wikipedia.random.
    Used for gathering and formatting written text representative of the 
Norwegian language,
    for use in training language models.
"""

# Create regex to filter the results
specialcharreg = re.compile(r'[^A-Za-zÆØÅæøå0-9.,-]+', re.IGNORECASE)
whitespacereg = re.compile(r' {2}', re.IGNORECASE)
punctuationreg = re.compile(r'[.]+', re.IGNORECASE)
shortsentencereg = re.compile(r'(</?s>)([a-zæøåA-ZÆØÅ0-9,\- ]{0,50})(</? 
s>)', re.IGNORECASE)
isbnreg = re.compile(r'(ISBN)([0-9- ]{7,21})', re.IGNORECASE)
nospaceaftertagreg = re.compile(r'(<s>([a-zæøåA-ZÆØÅ,-]))', re.IGNORECASE)


#  filter-methods for formatting the text
def nospeacialchar(wikicontent): return re.sub(specialcharreg, ' ', 
wikicontent)


def nodoublewhitespace(wikicontent): return re.sub(whitespacereg, ' ', 
wikicontent)


def faultysentence(wikicontent): return re.sub(shortsentencereg, '', 
wikicontent)


def inserttags(wikicontent): return re.sub(punctuationreg, ' </s>\n<s>', wikicontent)


def noemptylines(wikicontent): return "".join([s for s in wikicontent.splitlines(True) if s.strip("\r\n")])


def noisbn(wikicontent): return re.sub(isbnreg, '', wikicontent)


def nospaceaftertag(wikicontent): return re.sub(nospaceaftertagreg, '<s> ', wikicontent)


# We only want articles written in Norwegian
wikipedia.set_lang("no")

#  initialize different counters for counting duplicates and uniques
idlist = []
duplicatecount = 0
uniquecount: int = 0
showuniquecount = 0

# define number of pages to get
for x in range(0, 10001):
    try:
        randompages = wikipedia.random(1)
        for page in randompages:

            #  get wikipedia page
            wikipage = wikipedia.page(page)

            #  get page ID
            pageid = wikipage.pageid

            #  check for ID-duplicate
            if pageid not in idlist:

                #  add ID to list of gotten pages
                idlist.append(pageid)
                uniquecount += 1
                showuniquecount += 1

                #  on every tenth iteration, print current unique count
                if showuniquecount == 10:
                    print("Current unique page count:{0}".format(uniquecount))
                    showuniquecount = 0
                wikicontent = wikipage.content

                # filter the content using different regex-functions
                filteredcontent = \
                    faultysentence(
                        noemptylines(
                            nospaceaftertag(
                                faultysentence(
                                    inserttags(
                                        nodoublewhitespace(
                                            noisbn(
                                                nospeacialchar(
                                                    wikicontent))))))))
                print(filteredcontent)

                # Write operation to file
                with open("wikiscraping2.txt", "a", encoding="utf-8") as the_file:
                    the_file.write('<s> ' + filteredcontent)
                    the_file.close()
            else:
                duplicatecount += 1
                print("Duplicate! Current duplicate count:{0}".format(duplicatecount))

    #  catch exception of wikipedia not knowing which page is specified
    except wikipedia.DisambiguationError as e:
        print('DisambiguationError!')

        #  continue to next
        continue

    # catch exception
    except wikipedia.exceptions.PageError as d:
        print('Index error! (Page could not be found)')

        # continue to next
        continue

标签： pythonpython-3.xwikipediawikipedia-apimediawiki-api