PySentiBot.py

# coding: utf-8

# In[ ]:


# Dependencies
import tweepy
import json
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime

import os

import time

plt.style.use('fivethirtyeight')

# Import and Initialize Sentiment Analyzer
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
analyzer = SentimentIntensityAnalyzer()


consumer_key = os.getenv("bot_consumer_key")
consumer_secret = os.getenv("bot_consumer_secret")
access_token = os.getenv("bot_access_token")
access_token_secret = os.getenv("bot_access_token_secret")


# Setup Tweepy API Authentication
auth = tweepy.OAuthHandler(consumer_key, consumer_secret)
auth.set_access_token(access_token, access_token_secret)
api = tweepy.API(auth, parser=tweepy.parsers.JSONParser())


# In[ ]:


def parse_requests(tweet, tweet_dict=dict()):
    tweet_data = []
    tweet_id = tweet["id"]
    tweet_user = tweet["user"]["screen_name"]
    tweet_requests = []
    print(tweet_id)
    for mentions in tweet["entities"]["user_mentions"]:
        if (mentions["screen_name"] != "PySentimentBot"):
            tweet_requests.append(mentions["screen_name"])
    
    tweet_dict = {"id":tweet_id,"user":tweet_user,"analysis_requests":tweet_requests}
    return tweet_dict


# In[ ]:


def analyze_sentiments(recent_tweets, sentiment_results=list()):
    sentiment_results = []
    for tweet in recent_tweets:
        new_tweet = cleanse_tweet(tweet)
        sentiment_result = analyzer.polarity_scores(new_tweet["text"])
        sentiment_result.update({"tweet_id":new_tweet["id"]})
        sentiment_results.append(sentiment_result)    
    return sentiment_results


# In[ ]:


def remove_noise(tweet, category, key, result_tweet=dict()):
    try:
        result_tweet = tweet
        tweet_text = tweet.get("text")
        tweet_items = tweet.get("entities").get(category)
        for item in tweet_items:
            replace_str = item[key]
            tweet_text = tweet_text.replace(replace_str," ")
        result_tweet["text"] = tweet_text
    except TypeError:
        pass
    return result_tweet

def cleanse_tweet(tweet,result_tweet=dict()):
    result_tweet = tweet
    result_tweet = remove_noise(result_tweet,"user_mentions","screen_name")
    result_tweet = remove_noise(result_tweet,"urls","url")
    result_tweet = remove_noise(result_tweet,"media","url")
    result_tweet["text"] = result_tweet["text"].replace("@","")
    return result_tweet


# In[ ]:


def color_map(value):
    if(value >= 0): 
        return 'g'
    else:
        return 'r'

def plot_sentiments(title,sentiments):
    df = pd.DataFrame(sentiments)
    df = df.reset_index()
#     df.plot(kind="scatter",x="index",y="compound",marker="o")
    df.plot( 'index', 'compound', linestyle='-', marker='o',alpha=0.75)
    plt.ylabel("Sentiment score")
    plt.xlabel("Tweets")
    plt.title(title)
    
    filename = "SentimentAnalysis_of_"+title+".png"
    plt.savefig(filename)
    plt.show()
    
    return filename  


# In[ ]:


def scan_for_requests(since_tweet_id):
    
    total_tweets_so_far = 0
    
    last_tweet_id = since_tweet_id
    
    search_handle = "@PySentimentBot"

    results = api.mentions_timeline(since_tweet_id)
    
    print(f"Total results retrieved - {len(results)}")

    if(len(results) > 0):
        
        tweet_data = []
        print(results)

        for tweet in results:
            parsed_tweet = parse_requests(tweet)
            last_tweet_id = tweet['id']
            print(f"Parsed tweet - {tweet}")
            tweet_data.append(parsed_tweet)
        print(f"tweet data - {tweet_data}")
        
        for item in tweet_data:
            
            

            recent_tweets = []

            for analyze_request in item["analysis_requests"]:

                recent_tweets = api.user_timeline(analyze_request,count=200)
                
                print(f"{analyze_request} - {len(recent_tweets)}")
                
                if(len(recent_tweets) > 0):
                    sentiments = analyze_sentiments(recent_tweets)
                    print(sentiments)
                    sentiment_fig = plot_sentiments(analyze_request,sentiments)
                    text_status = f"{datetime.now()} - Thank you for your tweet @{item['user']}! Here is the sentiment analysis of {analyze_request}!"
                    api.update_with_media(filename=sentiment_fig,status=text_status,in_reply_to_status_id=item["id"])
                else:
                    text_status = f"{datetime.now()} - Thank you for your tweet @{item['user']}! Sorry, {analyze_request} has no tweets!" 
                    api.update_status(text_status)
                        
                    total_tweets_so_far = total_tweets_so_far + 1
                plt.show()
        return last_tweet_id
    else:
        return last_tweet_id


# In[ ]:


since_tweet_id = 937422923572400129

while True:
    time.sleep(30)
    since_tweet_id = scan_for_requests(since_tweet_id)