获取IEEE会议论文的标题和摘要

获取IEEE会议论文的标题和摘要 – 潘登同学的爬虫笔记

前几天接到导师的一个任务，要我去找找IEEE Transactions on Knowledge and Data Engineering期刊中与金融、企业有关的论文。起初我在IEEE官网不知所措，上了知网等很多论文网站，都很难将范围界定在这个期刊上。
后来我使用了谷歌学术的高级搜索，然后找了大概30篇文献，但是导师却说谷歌上的不全，让我去官网找，我又打开了官网。哪知道官网那网速真是感人，而且还不支持Ctrl+左键点开论文，每次点开论文看两眼，回去搜索结果又要等上一会儿。于是在研究了IEEE搜索结果及论文信息的页面后，写下了一个批量获取论文标题、论文摘要，及将其翻译成中文的全自动脚本。

打开IEEE的高级搜索

尽管爬虫再强大，他也不具备将人家论文数据库全弄下来的能力，还是要在官网的筛选引擎中进行初步筛选，点开Advanced search。 将关键词用引号括起来，以 OR 分割(记得加空格)。在出版标题中输入 IEEE Transactions on Knowledge and Data Engineering。

点击搜索。会进入一个结果界面，我们就能得到一个URL:

https://ieeexplore.ieee.org/search/searchresult.jsp?action=search&newsearch=true&matchBoolean=true&queryText=(%22All%20Metadata%22:%22financial%22%20OR%20%22All%20Metadata%22:%22finance%22%20OR%20%22All%20Metadata%22:%22trade%22%EF%BC%8C%22trading%22%20OR%20%22All%20Metadata%22:%22bank%22%20OR%20%22All%20Metadata%22:%22company%22%20OR%20%22All%20Metadata%22:%22enterprise%22%20OR%20%22All%20Metadata%22:%22management%22%20OR%20%22All%20Metadata%22:%22credit%22%20OR%20%22All%20Metadata%22:%22default%22%20OR%20%22All%20Metadata%22:%22risk%22%20OR%20%22All%20Metadata%22:%22asset%22%20OR%20%22All%20Metadata%22:%22bond%22%20OR%20%22All%20Metadata%22:%22stock%22%20OR%20%22All%20Metadata%22:%22equity%22%20OR%20%22All%20Metadata%22:%22volalitity%22%20OR%20%22All%20Metadata%22:%22futures%22%20OR%20%22All%20Metadata%22:%22share%22%20%22option%22%20OR%20%22All%20Metadata%22:%22return%22%20OR%20%22All%20Metadata%22:%22price%22%20OR%20%22All%20Metadata%22:%22pricing%22%20OR%20%22All%20Metadata%22:%22earning%22%20OR%20%22All%20Metadata%22:%22interest%22%20OR%20%22All%20Metadata%22:%22investment%22%20OR%20%22All%20Metadata%22:%22loan%22%20OR%20%22All%20Metadata%22:%22bankruptcy%22%20OR%20%22All%20Metadata%22:%22arbitrary%22)%20AND%20(%22Publication%20Title%22:IEEE%20Transactions%C2%A0on%C2%A0Knowledge%C2%A0and%20Data%20Engineering)&ranges=2020_2024_Year

但是三年的代码经验告诉我，这个URL不全，因为搜索结果是分页的，这个URL没有页码，所以我直接点下一页，获得另一个URL：

https://ieeexplore.ieee.org/search/searchresult.jsp?action=search&newsearch=true&matchBoolean=true&queryText=(%22All%20Metadata%22:%22financial%22%20OR%20%22All%20Metadata%22:%22finance%22%20OR%20%22All%20Metadata%22:%22trade%22%EF%BC%8C%22trading%22%20OR%20%22All%20Metadata%22:%22bank%22%20OR%20%22All%20Metadata%22:%22company%22%20OR%20%22All%20Metadata%22:%22enterprise%22%20OR%20%22All%20Metadata%22:%22management%22%20OR%20%22All%20Metadata%22:%22credit%22%20OR%20%22All%20Metadata%22:%22default%22%20OR%20%22All%20Metadata%22:%22risk%22%20OR%20%22All%20Metadata%22:%22asset%22%20OR%20%22All%20Metadata%22:%22bond%22%20OR%20%22All%20Metadata%22:%22stock%22%20OR%20%22All%20Metadata%22:%22equity%22%20OR%20%22All%20Metadata%22:%22volalitity%22%20OR%20%22All%20Metadata%22:%22futures%22%20OR%20%22All%20Metadata%22:%22share%22%20%22option%22%20OR%20%22All%20Metadata%22:%22return%22%20OR%20%22All%20Metadata%22:%22price%22%20OR%20%22All%20Metadata%22:%22pricing%22%20OR%20%22All%20Metadata%22:%22earning%22%20OR%20%22All%20Metadata%22:%22interest%22%20OR%20%22All%20Metadata%22:%22investment%22%20OR%20%22All%20Metadata%22:%22loan%22%20OR%20%22All%20Metadata%22:%22bankruptcy%22%20OR%20%22All%20Metadata%22:%22arbitrary%22)%20AND%20(%22Publication%20Title%22:IEEE%20Transactions%C2%A0on%C2%A0Knowledge%C2%A0and%20Data%20Engineering)&ranges=2020_2024_Year&highlight=true&returnFacets=ALL&returnType=SEARCH&matchPubs=true&pageNumber=2

此时我们能看到在Year之后出现了很多参数，这些应该都是默认参数不用改，但是最后一个是pageNumber=2这时候页码才是这个URL的关键。

环境准备

我们需要创建一个虚拟环境专门用于爬虫编写，因为在写爬虫的过程中经常需要调试，VScode支持的调试版本是py3.7以上，所以选择了比较熟悉的3.8。

conda create -n beautifulsoup_py38 python=3.8
conda activate beautifulsoup_py38
conda install requests
conda install pandas
conda install bs4
conda install lxml
conda install selenium

不同于静态网页，现在很多网页都用AJAX或者别的方法动态加载，如果用bs4最常用的方式，只会得到一串JS代码，这就是一个重大调整，后来我找到了selenium这个库，Selenium是一个用电脑模拟人操作浏览器网页，可以实现自动化，测试等。说白了这个方法很蠢，就是打开网页然后等着，等他加载完再爬…但是好在能全自动。

web = requests.get(url, headers = myHttpheader)
web.encoding = 'utf-8'   # important
soup = BeautifulSoup(web.text,'lxml')

使用selenium前除了下载库外，还需要下载浏览器驱动，我用的是Chrome浏览器，在设置里可以看到最新版本。我的是118版本。

• 114之前版本 http://chromedriver.storage.googleapis.com/index.html
• 116版本 https://edgedl.me.gvt1.com/edgedl/chrome/chrome-for-testing/116.0.5845.96/win64/chromedriver-win64.zip
• 117-119 https://googlechromelabs.github.io/chrome-for-testing/

下载好后，解压找到exe文件，

1. 将exe复制到 C:\Program Files\Google\Chrome\Application (与Chrome.exe在相同文件下)
2. 将exe复制到C:\Users\pdnbplus\anaconda3\envs\beautifulsoup_py38 (与python.exe在相同文件下)

配置环境变量:此电脑→右击属性→高级系统设置→环境变量→用户变量→Path→编辑→新建 C:\Program Files\Google\Chrome\Application\ 最后记得确认

申请百度密匙：
百度翻译开放平台：http://api.fanyi.baidu.com/api/trans/product/index

既然百度翻译需要验证密匙，我们就有必要先申请账号，获取密匙。申请完后点开发者信息就可以获取密匙。

完整爬虫过程

获取文章地址

在筛选界面随便点一篇文章，能看到URL是

https://ieeexplore.ieee.org/document/9942340

然后进入浏览器开发视图，看看链接上的href，能看到是/document/9942340/,那么我们就只需要收集所有href，然后拼接一下就能获取文章了。

因为文章标题的a标签的class是fw-bold,所有设置了一个等待，等待出现fw-bold这个class的时候才开始解析，最后将所有href放进一个列表，关闭浏览器。

import pandas as pd
import os
import time
import numpy as np
from bs4 import BeautifulSoup
from selenium import webdriver
from selenium.webdriver.support.wait import WebDriverWait #等待页面加载
from selenium.webdriver.common.by import By
from selenium.webdriver.support import expected_conditions as EC

# 设置浏览器选项
options = webdriver.ChromeOptions()
# options.add_argument('--headless') # 无头模式，不打开浏览器窗口

def get_urls(url):
    # 因为文章是js动态加载的，所以先要加载js加载的内容
    try:
        # 创建浏览器对象
        driver = webdriver.Chrome(options=options)
        driver.implicitly_wait(20)
        # 访问网页
        driver.get(url)
        driver.find_element_by_class_name('fw-bold')

        # 获取动态加载的网页内容
        dynamic_content = driver.page_source

        # 使用BeautifulSoup解析动态内容
        soup = BeautifulSoup(dynamic_content, 'html.parser')
        # web = requests.get(url, headers = myHttpheader)
        # web.encoding = 'utf-8'   # important
        # soup = BeautifulSoup(web.text,'lxml')
        # print(web.text)
        soup = soup.findAll('a',attrs={'class':'fw-bold'})

        # pattern = r'/document/\d+/'
        urls=[]
        for x in soup:
            urls.append(x['href'][:-1])
        urls = list(set(urls))
        print(len(urls))
    except:
        pass
    # 关闭浏览器
    driver.quit()
    return urls

# 阶段调试
if __name__ == '__main__':
    get_urls('https://ieeexplore.ieee.org/search/searchresult.jsp?action=search&newsearch=true&matchBoolean=true&queryText=(%22All%20Metadata%22:%22financial%22%20OR%20%22All%20Metadata%22:%22finance%22%20OR%20%22All%20Metadata%22:%22trade%22%EF%BC%8C%22trading%22%20OR%20%22All%20Metadata%22:%22bank%22%20OR%20%22All%20Metadata%22:%22company%22%20OR%20%22All%20Metadata%22:%22enterprise%22%20OR%20%22All%20Metadata%22:%22management%22%20OR%20%22All%20Metadata%22:%22credit%22%20OR%20%22All%20Metadata%22:%22default%22%20OR%20%22All%20Metadata%22:%22risk%22%20OR%20%22All%20Metadata%22:%22asset%22%20OR%20%22All%20Metadata%22:%22bond%22%20OR%20%22All%20Metadata%22:%22stock%22%20OR%20%22All%20Metadata%22:%22equity%22%20OR%20%22All%20Metadata%22:%22volalitity%22%20OR%20%22All%20Metadata%22:%22futures%22%20OR%20%22All%20Metadata%22:%22share%22%20%22option%22%20OR%20%22All%20Metadata%22:%22return%22%20OR%20%22All%20Metadata%22:%22price%22%20OR%20%22All%20Metadata%22:%22pricing%22%20OR%20%22All%20Metadata%22:%22earning%22%20OR%20%22All%20Metadata%22:%22interest%22%20OR%20%22All%20Metadata%22:%22investment%22%20OR%20%22All%20Metadata%22:%22loan%22%20OR%20%22All%20Metadata%22:%22bankruptcy%22%20OR%20%22All%20Metadata%22:%22arbitrary%22)%20AND%20(%22Publication%20Title%22:IEEE%20Transactions%C2%A0on%C2%A0Knowledge%C2%A0and%20Data%20Engineering)&ranges=2020_2024_Year&highlight=true&returnFacets=ALL&returnType=SEARCH&matchPubs=true&pageNumber=1')

看到如果输出25，那就是没问题了，因为默认一页有25篇文章

翻译函数

因为需要把英文的标题和摘要翻译成中文，所以先写一个翻译函数

import http.client
import hashlib
import urllib
import random
import json

appid = 'xxx'        # 填写你的appid
secretKey = 'xxx'    # 填写你的密钥
httpClient = None
url_baidu = '/api/trans/vip/translate'  # 通用翻译API HTTP地址

def translateBaidu(text, f='en', t='zh'):
    salt = random.randint(32768, 65536)
    sign = appid + text + str(salt) + secretKey
    sign = hashlib.md5(sign.encode()).hexdigest()
    url = url_baidu + '?appid=' + appid + '&q=' + urllib.parse.quote(text) + '&from=' + f + '&to=' + t + \
            '&salt=' + str(salt) + '&sign=' + sign
    try:
        httpClient = http.client.HTTPConnection('api.fanyi.baidu.com')
        httpClient.request('GET', url)
        # response是HTTPResponse对象
        response = httpClient.getresponse()
        result_all = response.read().decode("utf-8")
        data = json.loads(result_all)
        result = str(data['trans_result'][0]['dst'])
        return result
    except Exception as e:
            print (e)
    finally:
        if httpClient:
            httpClient.close()
# 阶段调试
if __name__ == '__main__':
    translateBaidu('i am happy')

看到如果输我很高兴,那么就没有问题

获取文章标题和摘要

根据前面找到的URL规律，对文章标题与摘要的标签进行解析得到文本内容，其中有一些文章没有doi代码，主要是书籍相关的部分，需要做一些异常处理。并进行翻译。

def get_info(urls):
    url1='https://ieeexplore.ieee.org'+urls
    # 创建浏览器对象
    driver = webdriver.Chrome(options=options)
    # 访问网页
    driver.get(url1)
    content = driver.page_source
    try:
        # 使用BeautifulSoup解析内容
        soup = BeautifulSoup(content, 'html.parser')
        # 显式等待：指定等待某个标签加载完毕
        wait=WebDriverWait(driver,5)
        wait.until(EC.presence_of_element_located((By.CLASS_NAME,'document-main')))

        tle = soup.find('h1',attrs={'class':"document-title text-2xl-md-lh"}).find('span').text
        title = tle.strip()
        abstract = soup.findAll('div',attrs={'class': 'u-mb-1'})[1].text

        zh_title = translateBaidu(title)
        zh_abstract = translateBaidu(abstract)
        try:
            doi = soup.find('div',attrs={'class': 'u-pb-1 stats-document-abstract-doi'}).find('a')['href']
            info = {'title':title,'标题':zh_title,'abstract':abstract,'摘要':zh_abstract,'doi':doi}
            driver.quit()
            return info
        except:
            info = {'title':title,'标题':zh_title,'abstract':abstract,'摘要':zh_abstract}
            driver.quit()
            return info
    except:
        driver.quit()

# 阶段调试
if __name__ == '__main__':
    url = '/document/10154753'
    get_info(urls=url)

最后将所有结合起来，写成一段脚本,为了减少难以避免的问题对文章收集的影响，我一页保存一次，但是这个官网实在比较慢，平均一页都需要10分钟。

if __name__ == '__main__':
    # set the work file directory
    path = r'C:/Users/pdnbplus/Documents/python全系列/网络爬虫/爬取IEEE文章/result'
    if not os.path.exists(path):
        print(path)
        os.mkdir(path)
        os.chdir(path)

    # Get the start urls.
    start_urls = 'https://ieeexplore.ieee.org/search/searchresult.jsp?action=search&newsearch=true&matchBoolean=true&queryText=(%22All%20Metadata%22:%22financial%22%20OR%20%22All%20Metadata%22:%22finance%22%20OR%20%22All%20Metadata%22:%22trade%22%EF%BC%8C%22trading%22%20OR%20%22All%20Metadata%22:%22bank%22%20OR%20%22All%20Metadata%22:%22company%22%20OR%20%22All%20Metadata%22:%22enterprise%22%20OR%20%22All%20Metadata%22:%22management%22%20OR%20%22All%20Metadata%22:%22credit%22%20OR%20%22All%20Metadata%22:%22default%22%20OR%20%22All%20Metadata%22:%22risk%22%20OR%20%22All%20Metadata%22:%22asset%22%20OR%20%22All%20Metadata%22:%22bond%22%20OR%20%22All%20Metadata%22:%22stock%22%20OR%20%22All%20Metadata%22:%22equity%22%20OR%20%22All%20Metadata%22:%22volalitity%22%20OR%20%22All%20Metadata%22:%22futures%22%20OR%20%22All%20Metadata%22:%22share%22%20%22option%22%20OR%20%22All%20Metadata%22:%22return%22%20OR%20%22All%20Metadata%22:%22price%22%20OR%20%22All%20Metadata%22:%22pricing%22%20OR%20%22All%20Metadata%22:%22earning%22%20OR%20%22All%20Metadata%22:%22interest%22%20OR%20%22All%20Metadata%22:%22investment%22%20OR%20%22All%20Metadata%22:%22loan%22%20OR%20%22All%20Metadata%22:%22bankruptcy%22%20OR%20%22All%20Metadata%22:%22arbitrary%22)%20AND%20(%22Publication%20Title%22:IEEE%20Transactions%C2%A0on%C2%A0Knowledge%C2%A0and%20Data%20Engineering)&ranges=2020_2024_Year&highlight=true&returnFacets=ALL&returnType=SEARCH&matchPubs=true&pageNumber='
    for i in range(10,21):
        info_all = [] 
        url_i = start_urls + str(i)
        top_urls = []
        try:
            top_urls = top_urls + get_urls(url_i)
        except Exception as e:
            print(i,e)
        for url in top_urls:
            time.sleep(random.randint(1,5)/10)
            info = get_info(url)
            if info and len(info)!=0:
                info_all = info_all + [info]           
        info_all = pd.DataFrame(info_all)
        info_all.to_csv(path+f'/article{i}.csv')