Twitter provides two types of API to access their data:
The reason why you would like to use streaming API:
Twitter streaming API can provide data through a streaming HTTP response. This is very similar to downloading a file where you read a number of bytes and store it to disk and repeat until the end of file. The only difference is this stream is endless. The only things that could stop this stream are:
This means that this process will be using the thread that it was launched from until it is stopped. In production, you should always start this in a different thread or process to make sure your software doesn't freeze until you stop the stream.
You will need four numbers from twitter development to start using streaming API. First, let's import some important libraries for dealing with twitter API, data analysis, data storage ... etc
import numpy as np
import pandas as pd
import tweepy
import matplotlib.pyplot as plt
import pymongo
import ipywidgets as wgt
from IPython.display import display
from sklearn.feature_extraction.text import CountVectorizer
import re
from datetime import datetime
%matplotlib inline
api_key = "yP0yoCitoUNgD63ebMerGyJaE" # <---- Add your API Key
api_secret = "kLO5YUtlth3cd4lOHLy8nlLHW5npVQgUfO4FhsyCn6wCMIz5E6" # <---- Add your API Secret
access_token = "259862037-iMXNjfL8JBApm4LVcdfwc3FcMm7Xta4TKg5cd44K" # <---- Add your access token
access_token_secret = "UIgh08dtmavzlvlWWukIXwN5HDIQD0wNwyn5sPzhrynBf" # <---- Add your access token secret
auth = tweepy.OAuthHandler(api_key, api_secret)
auth.set_access_token(access_token, access_token_secret)
api = tweepy.API(auth)
Connect to MongoDB and create/get a collection.
col = pymongo.MongoClient()["tweets"]["StreamingTutorial"]
col.count()
2251
We need a listener which should extend tweepy.StreamListener
class. There is a number of methods that you can extend to instruct the listener class to perform functionality. Some of the important methods are:
on_status(self, status)
: This will pass a status "tweet" object when a tweet is receivedon_data(self, raw_data)
: Called when any any data is received and the raw data will be passedon_error(self, status_code)
: Called when you get a response with code other than 200 (ok)class MyStreamListener(tweepy.StreamListener):
counter = 0
def __init__(self, max_tweets=1000, *args, **kwargs):
self.max_tweets = max_tweets
self.counter = 0
super().__init__(*args, **kwargs)
def on_connect(self):
self.counter = 0
self.start_time = datetime.now()
def on_status(self, status):
# Increment counter
self.counter += 1
# Store tweet to MongoDB
col.insert_one(status._json)
if self.counter % 1 == 0:
value = int(100.00 * self.counter / self.max_tweets)
mining_time = datetime.now() - self.start_time
progress_bar.value = value
html_value = """<span class="label label-primary">Tweets/Sec: %.1f</span>""" % (self.counter / max([1,mining_time.seconds]))
html_value += """ <span class="label label-success">Progress: %.1f%%</span>""" % (self.counter / self.max_tweets * 100.0)
html_value += """ <span class="label label-info">ETA: %.1f Sec</span>""" % ((self.max_tweets - self.counter) / (self.counter / max([1,mining_time.seconds])))
wgt_status.value = html_value
#print("%s/%s" % (self.counter, self.max_tweets))
if self.counter >= self.max_tweets:
myStream.disconnect()
print("Finished")
print("Total Mining Time: %s" % (mining_time))
print("Tweets/Sec: %.1f" % (self.max_tweets / mining_time.seconds))
progress_bar.value = 0
myStreamListener = MyStreamListener(max_tweets=100)
myStream = tweepy.Stream(auth = api.auth, listener=myStreamListener)
There are two methods to connect to a stream:
filter(follow=None, track=None, async=False, locations=None, stall_warnings=False, languages=None, encoding='utf8', filter_level=None)
firehose(count=None, async=False)
Firehose captures everything. You should make sure that you have connection speed that can handle the stream and you have the storage capacity that can store these tweets at the same rate. We cannot really use firehose for this tutorial but we'll be using filter
.
You have to specify one of two things to filter:
follow
: A list of user ID to follow. This will stream all their tweets, retweets, and others retweeting their tweets. This doesn't include mentions and manual retweets where the user doesn't press the retweet button.track
: A string or list of string to be used for filtering. If you use multiple words separated by spaces, this will be used for AND operator. If you use multiple words in a string separated by commas or pass a list of words this will be treated as OR operator.Note: track
is case insensitive.
I want to collect all tweets that contains any of these words:
This could be done with a string or a list. It is easier to to it with a list to make your code clear to read.
keywords = ["Jupyter",
"Python",
"Data Mining",
"Machine Learning",
"Data Science",
"Big Data",
"DataMining",
"MachineLearning",
"DataScience",
"BigData",
"IoT",
"#R",
]
# Visualize a progress bar to track progress
progress_bar = wgt.IntProgress(value=0)
display(progress_bar)
wgt_status = wgt.HTML(value="""<span class="label label-primary">Tweets/Sec: 0.0</span>""")
display(wgt_status)
# Start a filter with an error counter of 20
for error_counter in range(20):
try:
myStream.filter(track=keywords)
print("Tweets collected: %s" % myStream.listener.counter)
print("Total tweets in collection: %s" % col.count())
break
except:
print("ERROR# %s" % (error_counter + 1))
Finished Total Mining Time: 0:01:21.477351 Tweets/Sec: 1.2 Tweets collected: 100 Total tweets in collection: 2351
Now that we have stored all these tweets in a MongoDB collection, let's take a look at one of these tweets
col.find_one()
{'_id': ObjectId('56937d2e105f1970314720e2'), 'contributors': None, 'coordinates': None, 'created_at': 'Mon Jan 11 10:00:14 +0000 2016', 'entities': {'hashtags': [{'indices': [22, 27], 'text': 'Rã®æ³•å‰‡'}], 'symbols': [], 'urls': [], 'user_mentions': []}, 'favorite_count': 0, 'favorited': False, 'filter_level': 'low', 'geo': None, 'id': 686487772970942466, 'id_str': '686487772970942466', 'in_reply_to_screen_name': None, 'in_reply_to_status_id': None, 'in_reply_to_status_id_str': None, 'in_reply_to_user_id': None, 'in_reply_to_user_id_str': None, 'is_quote_status': False, 'lang': 'ja', 'place': None, 'retweet_count': 0, 'retweeted': False, 'source': '<a href="http://twitter.com/download/iphone" rel="nofollow">Twitter for iPhone</a>', 'text': '体力è½ã¡ã¦ãã¦ãŠã°ã•ã‚“ã¿ãŸã„ã«ãªã£ã¦ããŸã€‚\n#Rã®æ³•å‰‡', 'timestamp_ms': '1452506414059', 'truncated': False, 'user': {'contributors_enabled': False, 'created_at': 'Tue Aug 18 16:19:16 +0000 2015', 'default_profile': True, 'default_profile_image': False, 'description': '☮ 関ジャニ∞ & 山田涼介 & Justin Bieber & Benjamin Lasnier & Selena Gomez ☮', 'favourites_count': 1121, 'follow_request_sent': None, 'followers_count': 121, 'following': None, 'friends_count': 92, 'geo_enabled': True, 'id': 3318871652, 'id_str': '3318871652', 'is_translator': False, 'lang': 'en', 'listed_count': 0, 'location': 'The land of dreams', 'name': 'rena', 'notifications': None, 'profile_background_color': 'C0DEED', 'profile_background_image_url': 'http://abs.twimg.com/images/themes/theme1/bg.png', 'profile_background_image_url_https': 'https://abs.twimg.com/images/themes/theme1/bg.png', 'profile_background_tile': False, 'profile_banner_url': 'https://pbs.twimg.com/profile_banners/3318871652/1452436374', 'profile_image_url': 'http://pbs.twimg.com/profile_images/683964013558931456/Q1rx1s5b_normal.jpg', 'profile_image_url_https': 'https://pbs.twimg.com/profile_images/683964013558931456/Q1rx1s5b_normal.jpg', 'profile_link_color': '0084B4', 'profile_sidebar_border_color': 'C0DEED', 'profile_sidebar_fill_color': 'DDEEF6', 'profile_text_color': '333333', 'profile_use_background_image': True, 'protected': False, 'screen_name': 'Q2HpiJwCX1huBwf', 'statuses_count': 497, 'time_zone': None, 'url': None, 'utc_offset': None, 'verified': False}}
dataset = [{"created_at": item["created_at"],
"text": item["text"],
"user": "@%s" % item["user"]["screen_name"],
"source": item["source"],
} for item in col.find()]
dataset = pd.DataFrame(dataset)
dataset
created_at | source | text | user | |
---|---|---|---|---|
0 | Mon Jan 11 10:00:14 +0000 2016 | <a href="http://twitter.com/download/iphone" r... | 体力è½ã¡ã¦ãã¦ãŠã°ã•ã‚“ã¿ãŸã„ã«ãªã£ã¦ããŸã€‚\n#Rã®æ³•å‰‡ | @Q2HpiJwCX1huBwf |
1 | Mon Jan 11 10:09:26 +0000 2016 | <a href="http://twitter.com/download/android" ... | 皆ã«ãŠã°ã•ã‚“ã¨è¨€ã‚ã‚Œã¦ã†ã‚Œã—ãŒã£ã¦ã‚‹ #Rã®æ³•å‰‡ | @Tamutamu1017 |
2 | Mon Jan 11 10:00:10 +0000 2016 | <a href="http://trendkeyword.blog.jp/" rel="no... | ã€R.I.P】急上昇ワード「R.I.Pã€ã®ã¾ã¨ã‚é€Ÿå ± https://t.co/yi1yfC... | @pickword_matome |
3 | Mon Jan 11 10:00:10 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | #Rã®æ³•å‰‡ \nã©ã‚Œã‚‚ãŠã°ã•ã‚“è‡ã„ã‘ã‚Œã©ã‚„ã£ã±ã‚Šé»„色ãŒä¸€ç•ªã ãªã | @kakinotise |
4 | Mon Jan 11 10:00:10 +0000 2016 | <a href="http://bufferapp.com" rel="nofollow">... | The New Best Thing HP ATP - Vertica Big Data S... | @DataCentreNews1 |
5 | Mon Jan 11 10:00:11 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | IoT Now: l’Internet of Things è qui, ora https... | @datamanager_it |
6 | Mon Jan 11 10:00:11 +0000 2016 | <a href="http://trendkeyword.doorblog.jp/" rel... | 今話題ã®ã€ŒR.I.Pã€ã¾ã¨ã‚ https://t.co/VOc5cwK5hg #R.I.P ... | @buzz_wadai |
7 | Mon Jan 11 10:00:11 +0000 2016 | <a href="http://twitterfeed.com" rel="nofollow... | #oldham #stockport VIDEO: Snake thief hides py... | @Labour_is_PIE |
8 | Mon Jan 11 10:00:11 +0000 2016 | <a href="https://about.twitter.com/products/tw... | Las #startup pioneras de #machinelearning ofre... | @techreview_es |
9 | Mon Jan 11 10:00:12 +0000 2016 | <a href="http://www.linkedin.com/" rel="nofoll... | Lets talk about how to harness the power of ma... | @jansmit1 |
10 | Mon Jan 11 10:00:13 +0000 2016 | <a href="http://catalystfive.com" rel="nofollo... | Business Intelligence and Big Data Consulting ... | @Catalyst5Jobs |
11 | Mon Jan 11 10:00:13 +0000 2016 | <a href="http://twitter.com/NewsICT" rel="nofo... | [æƒ…å ±é€šä¿¡]2016å¹´å°åŒ—国際コンピューター見本市ãŒæ–°ã—ã„ä½ç½®ã¥ã‘ã¨æ–°ã—ã„展示ã§è£…ã„æ–°ãŸã«ï¼... | @NewsICT |
12 | Mon Jan 11 10:02:10 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | #bonplan Parties de Laser Quest entre amis à 2... | @Bons_Plans_ |
13 | Mon Jan 11 10:02:10 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | Parties de Laser Quest entre amis à 22.00€ au ... | @keepmymindfree |
14 | Mon Jan 11 10:02:10 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @jose_garde: Why a Simple Data Analytics St... | @martingeldish |
15 | Mon Jan 11 10:02:11 +0000 2016 | <a href="http://twitter.com/download/iphone" r... | 芸能人ã®äººãŸãã•ã‚“手ãŸãŸã笑ã„ã—ã¦ã‚‹ã‹ã‚‰ãŠã°ã•ã‚“ãŸãã•ã‚“ã«ãªã£ã¡ã‚ƒã†ã‚ˆwww\n\n#Rã®æ³•å‰‡ | @YK__0704 |
16 | Mon Jan 11 10:02:12 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | Découvrez le jeu Pure Mission entre amis à 22.... | @keepmymindfree |
17 | Mon Jan 11 10:02:12 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | #bonplan Parties de bowling pour 4 Ã #POINCY :... | @Bons_Plans_ |
18 | Mon Jan 11 10:02:12 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | 20 min de vol découverte ULM pour 1 ou 2 à 79.... | @CrationSiteWeb |
19 | Mon Jan 11 10:02:12 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @hortonworks: Paris is the city of love but... | @bigdataparis |
20 | Mon Jan 11 10:02:12 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @hynek: So #emacs / @spacemacs nerds: is th... | @fdiesch |
21 | Mon Jan 11 10:02:12 +0000 2016 | <a href="http://www.hootsuite.com" rel="nofoll... | .@QonexCyber founder member of @IoT_SF is orga... | @QonexCyber |
22 | Mon Jan 11 10:02:13 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | Parties de bowling pour 4 à #POINCY : 35.00€ a... | @keepmymindfree |
23 | Mon Jan 11 10:02:13 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | #bonplan 30 séances de Squash à #LISSES : 39.9... | @Bons_Plans_ |
24 | Mon Jan 11 10:02:13 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | App: ExZeus 2 – free to play https://t.co/ZT... | @UniversalConsol |
25 | Mon Jan 11 10:02:13 +0000 2016 | <a href="http://dlvr.it" rel="nofollow">dlvr.i... | #discount Parties de bowling pour 4 Ã #POINCY ... | @PromosPromos |
26 | Mon Jan 11 10:02:13 +0000 2016 | <a href="http://www.google.com/" rel="nofollow... | spiegel.de : Tier macht Sachen: Python beißt ... | @arminfischer_de |
27 | Mon Jan 11 10:09:02 +0000 2016 | <a href="http://twitter.com/download/android" ... | è‹¥ã„ã£ã¦ã„ã„ããˆâ€¦ã£ã¦ã‚ˆã言ã†w #Rã®æ³•å‰‡ | @naco75x |
28 | Mon Jan 11 10:09:17 +0000 2016 | <a href="http://twitter.com/download/iphone" r... | 最近ã®è‹¥ã„åã¯æœ€è¿‘使ã£ãŸ\n#Rã®æ³•å‰‡ | @K1224West |
29 | Mon Jan 11 10:09:19 +0000 2016 | <a href="http://twitter.com/download/android" ... | #Rã®æ³•å‰‡\n自分も若者ãªã®ã«ç¬‘ | @V6ZRRT7Q22BZ1cF |
... | ... | ... | ... | ... |
2221 | Mon Jan 11 10:29:40 +0000 2016 | <a href="http://201512291327-7430af.bitnamiapp... | https://t.co/BTAAq6HuuJ - pcgamer - #machinele... | @vinceyue |
2222 | Mon Jan 11 10:29:40 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @jose_garde: The Big Data Analytics Softwar... | @LJ_Blanchard |
2223 | Mon Jan 11 10:29:41 +0000 2016 | <a href="http://www.linkedin.com/" rel="nofoll... | What is data mining? Do you have to be a mathe... | @ednuwan |
2224 | Mon Jan 11 10:29:41 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @AgroKnow: A #BigData platform for the futu... | @albertspijkers |
2225 | Mon Jan 11 10:29:42 +0000 2016 | <a href="http://www.linkedin.com/" rel="nofoll... | Big Data: Is It A Tsunami, The New Oil, Or Sim... | @Summerlovegrove |
2226 | Mon Jan 11 10:29:43 +0000 2016 | <a href="https://www.jobfindly.com/php-jobs.ht... | Sr Software Engineer C Php Python Linux Jobs i... | @jobfindlyphpdev |
2227 | Mon Jan 11 10:29:43 +0000 2016 | <a href="http://twitter.com/download/iphone" r... | RT @bigdataparis: #Bigdata bang : un marché en... | @LifeIsWeb |
2228 | Mon Jan 11 10:29:43 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | Learn from the best professors in India .. cou... | @ashwaniapex |
2229 | Mon Jan 11 10:29:45 +0000 2016 | <a href="http://getsmoup.com" rel="nofollow">S... | RT @rebrandtoday: #startup or #rebrand -Buy Cr... | @SmartData_Fr |
2230 | Mon Jan 11 10:29:45 +0000 2016 | <a href="http://www.ajaymatharu.com/" rel="nof... | ¿Cómo será el futuro del Big Data? https://t.c... | @eduardogarsanch |
2231 | Mon Jan 11 10:29:46 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @jose_garde: 3 Ways to Transform Your Compa... | @LJ_Blanchard |
2232 | Mon Jan 11 10:29:46 +0000 2016 | <a href="http://publicize.wp.com/" rel="nofoll... | Woman Tries To Kiss Python, Gets Bitten In The... | @NAIJA_VIBEZ |
2233 | Mon Jan 11 10:29:46 +0000 2016 | <a href="http://www.itknowingness.com" rel="no... | RT @jose_garde: 3 Ways to Transform Your Compa... | @itknowingness |
2234 | Mon Jan 11 10:29:48 +0000 2016 | <a href="http://twitter.com/download/iphone" r... | RT @ErikaPauwels: Building a #BigData platform... | @impulsater |
2235 | Mon Jan 11 10:29:49 +0000 2016 | <a href="http://twitter.com/download/android" ... | En 2016 j'aimerais moins râler. #résolution. S... | @ce1ce2makarenko |
2236 | Mon Jan 11 10:29:51 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @jose_garde: How Marketing Can Be Better Au... | @LJ_Blanchard |
2237 | Mon Jan 11 10:29:51 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | Hey @Pontifex accueille ces réfugiées dans ta ... | @Atmosfive |
2238 | Mon Jan 11 10:29:51 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @Ubixr: .@GroupeLaPoste choisit le toulousa... | @The_Nextwork |
2239 | Mon Jan 11 10:29:56 +0000 2016 | <a href="http://twitterfeed.com" rel="nofollow... | Thanks @hackplayers Blade: un webshell en Pyth... | @Navarmedia |
2240 | Mon Jan 11 10:29:56 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @jose_garde: Which big data personality are... | @LJ_Blanchard |
2241 | Mon Jan 11 10:29:56 +0000 2016 | <a href="http://ifttt.com" rel="nofollow">IFTT... | Cybersecurity Forum tackles challenges with th... | @wulfsec |
2242 | Mon Jan 11 10:29:56 +0000 2016 | <a href="http://publicize.wp.com/" rel="nofoll... | Woman Tries To Kiss Python, Gets Bitten In The... | @Lola2Records |
2243 | Mon Jan 11 10:29:56 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @TeamAnodot: Join David Drai, CEO of Anodot... | @iottechexpo |
2244 | Mon Jan 11 10:29:57 +0000 2016 | <a href="http://getsmoup.com" rel="nofollow">A... | RT @rebrandtoday: #startup or #rebrand -Buy Cr... | @AI__news |
2245 | Mon Jan 11 10:29:57 +0000 2016 | <a href="http://www.twitter.com" rel="nofollow... | RT @Matthis__VERNON: "@Fred_Poquet Sans #confi... | @sibueta |
2246 | Mon Jan 11 10:29:57 +0000 2016 | <a href="https://social.zoho.com" rel="nofollo... | The right place for #BigData is #Cloud #Storag... | @TyroneSystems |
2247 | Mon Jan 11 10:29:59 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @PEBlanrue: Il a toujours le mot pour rire,... | @lesroisduring |
2248 | Mon Jan 11 10:29:58 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | RT @DigitalAgendaEU: €15 million for a #IoT so... | @ImproveNPA |
2249 | Mon Jan 11 10:29:59 +0000 2016 | <a href="http://twitter.com" rel="nofollow">Tw... | Neat IoT innovation https://t.co/atARX0m5Bj | @sherwinnovator |
2250 | Mon Jan 11 10:30:00 +0000 2016 | <a href="http://www.hubspot.com/" rel="nofollo... | Check out our #Mobile App Predicitions for 201... | @B60uk |
2251 rows × 4 columns
cv = CountVectorizer()
count_matrix = cv.fit_transform(dataset.text)
word_count = pd.DataFrame(cv.get_feature_names(), columns=["word"])
word_count["count"] = count_matrix.sum(axis=0).tolist()[0]
word_count = word_count.sort_values("count", ascending=False).reset_index(drop=True)
word_count[:50]
word | count | |
---|---|---|
0 | https | 1986 |
1 | co | 1907 |
2 | rt | 804 |
3 | de | 550 |
4 | rã®æ³•å‰‡ | 408 |
5 | iot | 374 |
6 | the | 358 |
7 | bigdata | 293 |
8 | 00 | 275 |
9 | data | 250 |
10 | in | 234 |
11 | python | 219 |
12 | to | 212 |
13 | au | 199 |
14 | of | 188 |
15 | lieu | 168 |
16 | réduction | 166 |
17 | big | 157 |
18 | on | 143 |
19 | is | 142 |
20 | and | 140 |
21 | for | 136 |
22 | analytics | 107 |
23 | le | 89 |
24 | via | 86 |
25 | you | 86 |
26 | thingsexpo | 86 |
27 | by | 85 |
28 | 2016 | 84 |
29 | snake | 80 |
30 | en | 76 |
31 | bowie | 75 |
32 | la | 74 |
33 | thief | 74 |
34 | video | 73 |
35 | m2m | 70 |
36 | jose_garde | 68 |
37 | 19 | 67 |
38 | david | 66 |
39 | with | 63 |
40 | how | 61 |
41 | it | 60 |
42 | will | 55 |
43 | un | 54 |
44 | amp | 53 |
45 | des | 53 |
46 | réparation | 53 |
47 | new | 52 |
48 | 39 | 52 |
49 | at | 51 |
def get_source_name(x):
value = re.findall(pattern="<[^>]+>([^<]+)</a>", string=x)
if len(value) > 0:
return value[0]
else:
return ""
dataset.source_name = dataset.source.apply(get_source_name)
source_counts = dataset.source_name.value_counts().sort_values()[-10:]
bottom = [index for index, item in enumerate(source_counts.index)]
plt.barh(bottom, width=source_counts, color="orange", linewidth=0)
y_labels = ["%s %.1f%%" % (item, 100.0*source_counts[item]/len(dataset)) for index,item in enumerate(source_counts.index)]
plt.yticks(np.array(bottom)+0.4, y_labels)
source_counts
Facebook 25 TweetDeck 37 RoundTeam 41 Hootsuite 46 twitterfeed 81 IFTTT 134 dlvr.it 200 Twitter Web Client 388 Twitter for Android 392 Twitter for iPhone 515 Name: source, dtype: int64