【337】Text Mining Using Twitter Streaming API and Python

Reference: An Introduction to Text Mining using Twitter Streaming API and Python

Reference: How to Register a Twitter App in 8 Easy Steps

• Getting Data from Twitter Streaming API
• Reading and Understanding the data
• Mining the tweets

Key Methods:

• Map()
• Lambda()
• Set()
• Pandas.DataFrame()
• matplotlib

---

1. Getting Data from Twitter Streaming API

twitter_streaming.py, this file is used to extract information from Twitter.

#Import the necessary methods from tweepy library
from tweepy.streaming import StreamListener
from tweepy import OAuthHandler
from tweepy import Stream

#Variables that contains the user credentials to access Twitter API 
access_token = "ENTER YOUR ACCESS TOKEN"
access_token_secret = "ENTER YOUR ACCESS TOKEN SECRET"
consumer_key = "ENTER YOUR API KEY"
consumer_secret = "ENTER YOUR API SECRET"


#This is a basic listener that just prints received tweets to stdout.
class StdOutListener(StreamListener):

    def on_data(self, data):
        print(data)
        return True

    def on_error(self, status):
        print(status)


if __name__ == '__main__':

    #This handles Twitter authetification and the connection to Twitter Streaming API
    l = StdOutListener()
    auth = OAuthHandler(consumer_key, consumer_secret)
    auth.set_access_token(access_token, access_token_secret)
    stream = Stream(auth, l)

    #This line filter Twitter Streams to capture data by the keywords: 'python', 'javascript', 'ruby'
    stream.filter(track=['python', 'javascript', 'ruby'])

You can use the following command to store information in the specific file. (By CMD)

python twitter_streaming.py > twitter_data.txt

Then we will get the information from the above text file and store them in JSON format.

import json
tweets_data_path = r"..	witter_data.txt"
tweets_data = []
tweets_file = open(tweets_data_path, "r")
for line in tweets_file:
	try:
		tweet = json.loads(line)
		tweets_data.append(tweet)
	except:
		continue

Data are stored in tweets_data, and we can get the specific information by the following scripts.

Reference: python JSON only get keys in first level

# get the text content, language from the specific tweets
num = 0
for tweet in tweets_data:
	num += 1
	if num == 10:
		break
	else:
		tweet_text = tweet["text"]
		tweet_lang = tweet["lang"]
		print(str(num))
		print(tweet_lang)
		print(tweet_text)
		print()

# get all the keys from json
tweets_data[0].keys()

2. Reading and Understanding the data

Now we can also get the specific key by list(), map() and lambda() with the following scripts.

Reference: Python中map与lambda的结合使用

>>> a = list(map(lambda tweet: tweet['text'], tweets_data))
>>> len(a)
1633
>>> a[0]
'RT @neet_se: 案件数って点だけならJavaがダントツ、つまり仕事に繋がりやすい。https://t.co/rqxp…'

Or we can also use set() method to get the unique values of the list.

Reference: Python set() 函数

Reference: Python统计列表中的重复项出现的次数的方法

>>> langs = list(map(lambda tweet: tweet['lang'], tweets_data))
>>> len(langs)
1633
>>> set(langs)
{'zh', 'de', 'es', 'et', 'th', 'cy', 'ru', 'in', 'lt', 'pt', 'tl', 'en', 'it', 'ja', 'ro', 'fa', 'pl', 'fr', 'ht', 'ar', 'tr', 'ca', 'cs', 'und', 'da'}

Next, we will structure the tweets data into a pandas DataFrame to simplify the data manipulation.

>>> import pandas as pd
>>> tweets = pd.DataFrame()
>>> tweets['text'] = list(map(lambda tweet: tweet['text'], tweets_data))
>>> tweets['lang'] = list(map(lambda tweet: tweet['lang'], tweets_data))
>>> tweets['country'] = list(map(lambda tweet: tweet['place']['country'] if tweet['place'] != None else None, tweets_data))
>>> tweets['lang'].value_counts()
en     1119
ja      278
es      113
pt       36
und      26
...

Next, we will use matplotlib to create a chart describing the Top 5 languages in which the tweets were written.

>>> tweets_by_lang = tweets['lang'].value_counts()

>>> import matplotlib.pyplot as plt
>>> fig, ax = plt.subplots()
>>> ax.tick_params(axis='x', labelsize=15)
>>> ax.tick_params(axis='y', labelsize=10)
>>> ax.set_xlabel('Languages', fontsize=15)
Text(0.5, 0, 'Languages')
>>> ax.set_ylabel('Number of tweets' , fontsize=15)
Text(0, 0.5, 'Number of tweets')
>>> ax.set_title('Top 5 languages', fontsize=15, fontweight='bold')
Text(0.5, 1.0, 'Top 5 languages')
>>> tweets_by_lang[:5].plot(ax=ax, kind='bar', color='red')
<matplotlib.axes._subplots.AxesSubplot object at 0x00000189B635D630>
>>> plt.show()

Next, we will create a chart describing the Top 5 countries from which the tweets were sent.

>>> tweets_by_country = tweets['country'].value_counts()

>>> fig, ax = plt.subplots()
>>> ax.tick_params(axis='x', labelsize=15)
>>> ax.tick_params(axis='y', labelsize=10)
>>> ax.set_xlabel('Countries', fontsize=15)
Text(0.5, 0, 'Countries')
>>> ax.set_ylabel('Number of tweets' , fontsize=15)
Text(0, 0.5, 'Number of tweets')
>>> ax.set_title('Top 5 countries', fontsize=15, fontweight='bold')
Text(0.5, 1.0, 'Top 5 countries')
>>> tweets_by_country[:5].plot(ax=ax, kind='bar', color='blue')
<matplotlib.axes._subplots.AxesSubplot object at 0x00000189BA6038D0>
>>> plt.show()

3. Mining the tweets

Out main goals in these text mining tasks are: compare the popularity of Python, Ruby and Javascript programming languages and to retrieve programming tutorial links. We will do this in 3 steps:

• We will add tags to our tweets DataFrame in order to be able to manipulate the data easily.
• Target tweets that have "programming" or tutorial" keywords.
• Extract links from the relevant tweets.

Adding Python, Ruby, and Javascript tags

First, we will create a function that checks if a specific keyword is present in a text. We will do this by using regular expression (正则表达式).

Python provides a library for regular expression called re. We will start by importing this library.

Next, we will create a function called word_in_text(word, text). This function return True if a word is found in text, otherwise it returns False.

>>> import re
>>> def word_in_text(word, text):
	word = word.lower()
	text = text.lower()
	match = re.search(word, text)
	if match:
		return True
	return False

Next, we will add 3 columns to our tweets DataFrame by pandas.DataFrame.apply().

>>> tweets['python'] = tweets['text'].apply(lambda tweet: word_in_text('python', tweet))
>>> tweets['ruby'] = tweets['text'].apply(lambda tweet: word_in_text('ruby', tweet))
>>> tweets['javascript'] = tweets['text'].apply(lambda tweet: word_in_text('javascript', tweet))

We can calculate the number of tweets for each programming language by pandas.Series.value_counts as follows:

>>> print(tweets['python'].value_counts()[True])	       
447
>>> print(tweets['ruby'].value_counts()[True])	       
529
>>> print(tweets['javascript'].value_counts()[True])	       
275

We can make a simple comparison chart by executing the following:

>>> prg_langs = ['python', 'ruby', 'javascript']  
>>> tweets_by_prg_lang = [tweets['python'].value_counts()[True], tweets['ruby'].value_counts()[True], tweets['javascript'].value_counts()[True]]     
>>> x_pos = list(range(len(prg_langs)))
>>> width = 0.8       
>>> fig, ax = plt.subplots()  
>>> plt.bar(x_pos, tweets_by_prg_lang, width, alpha=1, color='g')	       
<BarContainer object of 3 artists>
>>> # Setting axis labels and ticks       
>>> ax.set_ylabel('Number of tweets', fontsize=15)       
Text(0, 0.5, 'Number of tweets')
>>> ax.set_title('Ranking: python vs. javascript vs. ruby (Raw data)', fontsize=10, fontweight='bold')       
Text(0.5, 1.0, 'Ranking: python vs. javascript vs. ruby (Raw data)')
>>> ax.set_xticks([p + 0.4 * width for p in x_pos])      
[<matplotlib.axis.XTick object at 0x00000189BA5D1F28>, <matplotlib.axis.XTick object at 0x00000189BA603D30>, <matplotlib.axis.XTick object at 0x00000189BA5D15F8>]
>>> ax.set_xticklabels(prg_langs)       
[Text(0, 0, 'python'), Text(0, 0, 'ruby'), Text(0, 0, 'javascript')]
>>> plt.grid()       
>>> plt.show()

This shows, that the keyword ruby is the most popular, followed by python then javascript. However, the tweets DataFrame contains information about all tweets that contains one of the 3 keywords and doesn't restrict the information to the programming languages. For example, there are a lot of tweets that contains the keyword ruby and that are related to a political scandal Rubygate. In the next section, we will filter the tweets and re-run the analysis to make a more accurate comparison.

Targeting relevant tweets

We are interested in targeting tweets that are related to programming languages. Such tweets often have one of the 2 keywords: "programming" or "tutorial". We will create 2 additional columns to our tweets DataFrame where we will add this information.

>>> tweets['programming'] = tweets['text'].apply(lambda tweet: word_in_text('programming', tweet))
>>> tweets['tutorial'] = tweets['text'].apply(lambda tweet: word_in_text('tutorial', tweet))

We will add an additional column called relevant that take value True if the tweet has either "programming" or "tutorial" keyword, otherwise it takes value False.

>>> tweets['relevant'] = tweets['text'].apply(lambda tweet: word_in_text('programming', tweet) or word_in_text('tutorial', tweet))

We can print the counts of relevant tweet by executing the commands below.

>>> print(tweets['programming'].value_counts()[True])       
55
>>> print(tweets['tutorial'].value_counts()[True])       
22
>>> print(tweets['relevant'].value_counts()[True])  
74

We can compare now the popularity of the programming languages by executing the commands below.

tweets[tweets['relevant'] == True]['python'] # 将 relevant 为 True 的索引对应 Python 组成一个新的列

>>> print(tweets[tweets['relevant'] == True]['python'].value_counts()[True])       
31
>>> print(tweets[tweets['relevant'] == True]['ruby'].value_counts()[True])
8
>>> print(tweets[tweets['relevant'] == True]['javascript'].value_counts()[True])   
11

Python is the most popular with a count of 31, followed by javascript by a count of 11, and ruby by a count of 185. We can make a comparison

>>> tweets_by_prg_lang = [tweets[tweets['relevant'] == True]['python'].value_counts()[True],
			  tweets[tweets['relevant'] == True]['ruby'].value_counts()[True],
			  tweets[tweets['relevant'] == True]['javascript'].value_counts()[True]] 
>>> x_pos = list(range(len(prg_langs)))
>>> width = 0.8
>>> fig, ax = plt.subplots()
>>> plt.bar(x_pos, tweets_by_prg_lang, width,alpha=1,color='g')
<BarContainer object of 3 artists>
>>> ax.set_ylabel('Number of tweets', fontsize=15)
Text(0, 0.5, 'Number of tweets')
>>> ax.set_title('Ranking: python vs. javascript vs. ruby (Relevant data)', fontsize=10, fontweight='bold')
Text(0.5, 1.0, 'Ranking: python vs. javascript vs. ruby (Relevant data)')
>>> ax.set_xticks([p + 0.4 * width for p in x_pos])
[<matplotlib.axis.XTick object at 0x00000189B6E9E128>, <matplotlib.axis.XTick object at 0x00000189B430F9E8>, <matplotlib.axis.XTick object at 0x00000189B430F5C0>]
>>> ax.set_xticklabels(prg_langs) 
[Text(0, 0, 'python'), Text(0, 0, 'ruby'), Text(0, 0, 'javascript')]
>>> plt.grid()
>>> plt.show()

Extracting links from the relevants tweets

Now that we extracted the relevant tweets, we want to retrieve links to programming tutorials. We will start by creating a function that uses regular expressions for retrieving link that start with "http://" or "https:" from a text. This function will return the url if found, otherwise it returns an empty string.

>>> def extract_link(text):
	regex = r'https?://[^s<>"]+|www.[^s<>"]+'
	match = re.search(regex, text)
	if match:
		return match.group()
	return ''

Next, we will add a column called link to our tweets DataFrame. This column will contain the urls information.

>>> tweets['link'] = tweets['text'].apply(lambda tweet: extract_link(tweet))

Next, we will create a new DataFrame called tweets_relevant_with_link. This DataFrame is a subset of tweets DataFrame and contains all relevant tweets that have a link.

将原有 DataFrame 进行截取。

>>> tweets_relevant = tweets[tweets['relevant'] == True]       
>>> tweets_relevant_with_link = tweets_relevant[tweets_relevant['link'] != '']

We can now print out all links for python, ruby, and javascript by executing the commands below:

>>> print(tweets_relevant_with_link[tweets_relevant_with_link['python'] == True]['link'])       
40      https://t.co/zoAgyQuMAZ
105     https://t.co/ogaPbuIbEW
274     https://t.co/y4sUmovFOn
329     https://t.co/A030fqWeWA
339     https://t.co/LaaVc5T2rQ
391     https://t.co/8bYvlziCZb
413     https://t.co/8bYvlziCZb
436     https://t.co/EByqxT1qyN
444     https://t.co/8bYvlziCZb
445     https://t.co/5Jujg6h31B
462     https://t.co/UrFHlOaJYf
476     https://t.co/5Jujg6h31B
477     https://t.co/EByqxT1qyN
589     https://t.co/UrFHlOaJYf
603     https://t.co/5Jujg6h31B
822     https://t.co/Oc21FrzQc5
1060    https://t.co/qOAIuKfyD0
1097    https://t.co/qOAIuKfyD0
1248    https://t.co/V3ZNKuYsK7
1278    https://t.co/qOAIuKfyD0
1411    https://t.co/szHRHavQKy
1594    https://t.co/X6KWMlzlv6
Name: link, dtype: object
>>> print(tweets_relevant_with_link[tweets_relevant_with_link['ruby'] == True]['link'])	       
782     https://t.co/JgY40r2NSo
833     https://t.co/JgY40r2NSo
1177    https://t.co/xycOG3ndi9
1254    https://t.co/xycOG3ndi9
1293    https://t.co/LMHW050TGs
1328    https://t.co/SS4DzEnSBZ
1393    https://t.co/NZlUce5Ne8
1619    https://t.co/e4nwrn3N2j
Name: link, dtype: object
>>> print(tweets_relevant_with_link[tweets_relevant_with_link['javascript'] == True]['link'])     
130     https://t.co/AbJFaSI0B8
286     https://t.co/7dNBIsQ5Gq
467     https://t.co/3YIK588j8t
471     https://t.co/vjBJWWzvfv
830     https://t.co/T4mUjwUcgL
1093    https://t.co/wvLZLjuVKF
1180    https://t.co/luxL2qbxte
1526    https://t.co/G3ZTFL0RKv
Name: link, dtype: object