As part of a small school project, I have collected tweets from a few world cup games over the past month. I was using the 'hash flags' that twitter provided; so if I was grabbing tweets for Brazil vs Colombia I would be searching for '#bra or #col'.
I didn't store all of the information for each tweet, instead I just stored:
- time the tweet was created
- text from the tweet
- language of the tweet
- hashtags
- user location
- user utc offset
- user timezone
I have about 700k tweets for each game.
Here is something I through together really quickly that demonstrates the volume of tweets based on which hashtags were in the tweet.
The school project is to use this data to preform some sort of machine learning. Originally I wanted to use the tweet volume to identify potential goals (any spike in the data would be a potential positive), and do some sentiment analysis on each team's tweets to identify the false positives. The problem with this is that I will have to label the tweets myself.
tldr; I need to use this data to explore some concept in ML or NLP but I don't know what to do.
[link][2 comments]
