***** CITATION *****

Please cite our paper as follows, when you are using our dataset. Our paper is under review right now:

@unpublished{varol2017early,
  Author = {Onur Varol and Emilio Ferrara and Filippo Menczer and Alessandro Flammini},
  Note = {Under review},
  Title = {Early Detection of Artificially Promoted Memes on Social Media},
  Url = {}
  Year = {2017},
}

**** FORMAT ****

Information about promoted and organics trends used in this project available in the index file: 
`trend_information.dat`. We also provide `tweet-id` for each tweets used in our experiments and `timeseries` of features for each time window.

trend_information.dat:
  Format: {trend_name} \t {label} \t {trending_tstamp}
  * {trend-name} is the identifier of different trends. Information about each trend can be found in the index file.  
  * label: 1 (organic), 0 (promoted)
  * trending_tstamp provides a UNIX timestamp of the time particular hashtag is trended.

tweet-ids/tweet-ids_{trend-name}.dat.gz:
  Format: Each line contains a single tweet-id
  * {trend-name} is the identifier of different trends. Information about each trend can be found in the index file.

timeseries/timeseries
  Format: {feature-name} \t {timeseries}
  * {feature-name} represent different features extracted for a given trend
  * {timeseries} contain 120 comma seperated numerical value. Trending point locates at index 60.


---
Onur Varol, ovarol@indiana.edu
March 15, 2017
