Assignment for the Information retrieval course, Pisa 2011-12
The task was to capture and process a JSON collection of millions of tweets.
There is space in the machine so copy whatever you need.
In case you remove accidentally something mail me [email protected]
These are very simple scripts so if you need any field or data just tell me.
Take into account that I gzip'ed most of the files for speed porpuses. Some cases, like the tweet text, are not immediately gzip.
Comprensive list:
separate_users.py: transform the X.userinfo file into a (id, screen_name)
input: data/TweetTwitter-20110912_160840.usersinfo.gz
output: data/TweetTwitter-20110912_160840.usersinfo.cleaned.gz
parse_tweets.py: Read the Twitter dump and separete the data into a list of
tweets for every user (id, [tweet id]) and the first cleaned tweets (tweet id,
text). This first cleaning only remove newlines and alike.
The head_10 is an example with only the first ten lines of the original file
input: data/TweetTwitter-20110912_160840.tweet.head_10.gz
output: data/TweetTwitter-20110912_160840.tweet.head_10.tweet_list.gz
output: data/TweetTwitter-20110912_160840.tweet.head_10.text
input: data/TweetTwitter-20110912_160840.tweet.gz
output: data/TweetTwitter-20110912_160840.tweet.tweet_list.gz
output: data/TweetTwitter-20110912_160840.tweet.text
clean_tweets.py Almost useless right now, just removes the hashes. But this
gives a good framework if we want to apply heuristics and expand the hash tag.
TweetAnnotation.java (this file is found in the ~/src directory)
Reads the whole cleaned file and queries TAGME.
The output format is (tweet_id, [# Annotation])
input: /l/disc3/home/ir2011/paurullan/data/TweetTwitter-20110912_160840.tweet.head_10.clean
output: /l/disc3/home/ir2011/paurullan/data/TweetTwitter-20110912_160840.tweet.head_10.annotation