《演员请就位2》弹幕的情感倾向分析

最近小编的娱乐公众号被《演员请就位2》刷屏了,这部综艺的从开播开始导演的热搜话题就一直不断,我们用 Python 分析一下这部综艺的视频弹幕看看大家都在吐糟些什么。

弹幕抓取

在腾讯视频打开最新的第 8 期的上下两期,在 Network 面板中搜索【danmu】,找到弹幕的链接 (https://mfm.video.qq.com/danmu?otype=json….)

分析其中的请求参数可以发现只有 timestamp 参数在以每次 30 的数字递增,盲猜一波应该是视频每 30 秒获取一次弹幕包,其他的请求参数可以保持不变

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
import csv
import requests
import json
import time
from pathlib import Path

def danmu():
    headers = {
        'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/86.0.4240.80 Safari/537.36'
    }
    # 弹幕链接,视频长度(秒)
    urls = [['https://mfm.video.qq.com/danmu?otype=json&callback=&target_id=6208914107%26vid%3Do0035t7199o&session_key=63761%2C673%2C1606144955&timestamp={}&_=1606144949402', 7478],
            ['https://mfm.video.qq.com/danmu?otype=json&callback=&target_id=6208234802%26vid%3Da00352eyo25&session_key=111028%2C1191%2C1606200649&timestamp={}&_=1606200643186', 8610]]

    for url in urls:
        for page in range(15, url[1], 30):
            u = url[0].format(page)
            html = requests.get(u, headers=headers)
            result = json.loads(html.text, strict=False)
            time.sleep(1)
            danmu_list = []
            # 遍历获取目标字段
            for i in result['comments']:
                content = i['content']  # 弹幕内容
                danmu_list.append([content])
                print(content)
            csv_write(danmu_list)

def csv_write(tablelist):
    tableheader = ['弹幕内容']
    csv_file = Path('danmu.csv')
    not_file = not csv_file.is_file()
    with open('danmu.csv', 'a', newline='', errors='ignore') as f:
        writer = csv.writer(f)
        if not_file:
            writer.writerow(tableheader)
        for row in tablelist:
            writer.writerow(row)

抓到了 7W+ 的弹幕,文件为 3M 大小

情感分析

抓取到弹幕后,用腾讯云的情感分析 API 分析弹幕的情感倾向是正面的还是负面的亦或是中性情感

参考腾讯云 https://cloud.tencent.com/document/sdk/Python 页面获取 SecretId 和 SecretKey 安全凭证,用 pip install tencentcloud-sdk-python 安装腾讯云的 SDK,遇到证书错误时用 sudo "/Applications/Python 3.6/Install Certificates.command" 命令安装证书

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
from tencentcloud.common import credential
from tencentcloud.common.profile.client_profile import ClientProfile
from tencentcloud.common.profile.http_profile import HttpProfile
from tencentcloud.common.exception.tencent_cloud_sdk_exception import TencentCloudSDKException
from tencentcloud.nlp.v20190408 import nlp_client, models

import ssl
ssl._create_default_https_context=ssl._create_unverified_context

def nlp(text):
    try:
        cred = credential.Credential("xxx", "xxx")
        httpProfile = HttpProfile()
        httpProfile.endpoint = "nlp.tencentcloudapi.com"

        clientProfile = ClientProfile()
        clientProfile.httpProfile = httpProfile
        client = nlp_client.NlpClient(cred, "ap-guangzhou", clientProfile)

        req = models.SentimentAnalysisRequest()
        params = {
            "Text": text,
            "Mode": "3class"
        }
        req.from_json_string(json.dumps(params))

        resp = client.SentimentAnalysis(req)
        sentiment = {'positive': '正面', 'negative': '负面', 'neutral': '中性'}
        return sentiment[resp.Sentiment]
    except TencentCloudSDKException as err:
        print(err)

示例结果

导演好感度

对于频频上热搜的导演们观众对他们的感官是怎么样的,将情感分析结果转换成大家对各个导演评价的百分比,并用 pyecharts 制作成图表

弹幕中对赵薇的负面评价达到 30%,尔冬升、赵薇、郭敬明的正面评价都差不多在 46% 左右,主持人大鹏的正面评价居然是最高的,达到 59%,赵薇的弹幕量最多、陈凯歌弹幕数量是第二个,尔冬升的弹幕量不到 2000

弹幕词云

将弹幕词云化,看看大家都在吐槽写什么

第一眼就看到了的秋裤两个字

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
def ciyun():
    with open('danmu.csv') as f:
        with open('ciyun.txt', 'a') as ciyun_file:
            csv_reader = csv.reader(f)
            for row in csv_reader:
                ciyun_file.write(row[0])

    # 构建并配置词云对象w
    w = wordcloud.WordCloud(width=1000,
                            height=700,
                            background_color='white',
                            font_path="/System/Library/fonts/PingFang.ttc",
                            collocations=False,
                            stopwords={'的', '了','啊','我','很','是','好','这','都','不'})

    
    f = open('ciyun.txt', encoding='utf-8')
    txt = f.read()
    txtlist = jieba.lcut(txt)
    result = " ".join(txtlist)
    
    w.generate(result)

    w.to_file('演员请就位2.png')

总结

腾讯视频弹幕的抓取比较简单,每隔 30 秒发送一次请求获取弹幕包。有兴趣的朋友可以尝试其他视频网站的弹幕抓取,一起努力进步天天向上

示例代码:《演员请就位2》弹幕的情感倾向分析

Python Geek Tech wechat
欢迎订阅 Python 技术,这里分享关于 Python 的一切。