twitter_conversations.py

import sys
import json
from operator import itemgetter
import networkx as nx

def usage():
    print("Usage")
    print("python {} <filename>".format(sys.argv[0]))

if __name__ == "__main__":
    if len(sys.argv) != 2:
        usage()
        sys.exit(1)

    fname = sys.argv[1]
    with open(fname) as f: #takes in a jsonl file of tweets as input
        graph = nx.DiGraph()
        for line in f:
            tweet = json.loads(line)
            if 'id' in tweet:
                graph.add_node(tweet['id'],
                                tweet=tweet['text'],
                                author=tweet['user']['screen_name'],
                                created_at=tweet['created_at'])
                if tweet['in_reply_to_status_id']:
                    reply_to = tweet['in_reply_to_status_id']
                    if reply_to in graph and tweet['user']['screen_name'] != graph.node[reply_to]['author']:  #if the user is not replying to themselves
                        graph.add_edge(tweet['in_reply_to_status_id'], tweet['id'])
        #Print some basic stats

        print(nx.info(graph))

        #Find most replied tweet (most indegree)
        sorted_replied = sorted(graph.degree(), key=itemgetter(1), reverse=True)

        most_replied_id, replies = sorted_replied[0]
        print("Most replied tweet ({} replies:".format(replies))
        print(graph.node[most_replied_id])

        #Find longest conversation (longest path)
        print("Longest discussion:")
        longest_path = nx.dag_longest_path(graph)
        for tweet_id in longest_path:
            node = graph.node[tweet_id]
            print("{} (by {} at {})".format(node['tweet'], node['author'], node['created_at']))