An overview of the AAS225 tweet stream

Here are a few features of the tweets from the 225th meeting of the American Astronomical Society.

The page is updated infrequently.

Time range (Pacific): to

There have been tweets and re-tweets made by people, of which only sent retweets ().

The most retweeted tweets

The retweet numbers are close to the values that Twitter has for the tweets, which means that I can't have messed up too much (I wouldn't expect perfect agreement since Twitter does not guarantee that any search will return all matching tweets; also, the Twitter values are updated whereas mine are fixed until the next set of tweets gets analyzed).

A look at what's popular

The most re-tweeted accounts.
Twitter account Number of re-tweets
The accounts that are mentioned the most.
Twitter account Number of mentions

It is not really that surprising that the most re-tweeted accounts also appear in the most-mentioned table.

The accounts that are replied to the most (this excludes retweets)
Twitter account Number of replies received
The accounts that have replied to the most tweets (this excludes retweets)
Twitter account Number of replies made
What are the most-used hash tags?
HashTag Number of mentions % of tweets

Note that #aas225, #hackaas, and #aasviz have an advantage here since they match the search terms (using a case insensitive match).

What URLs are mentioned the most?
URL Number of mentions

NOTE: unfortunately the URL http://www.colorado.edu/news/releases/2014/06/04/astronomers-discover-first-thorne-żytkow-object-bizarre-type-hybrid-star is not included in the analysis since the code doesn't like the pesky ż character; I do not think it has been referenced enough to make it into the table above, but can't prove that at this time.

Note that I have attempted to follow the twisty-turny maze of URL redirects and link shortners that are prevelant in the TwitterVerse to get at the original URL (so if the same page has different links to it then they will be combined). Since there are some common query strings seen in URLs on Twitter which can lead to "missing" counts, I removed the following query terms for most URLS: utm_source, utm_medium, utm_campaign, goback and cid. For YouTube links I dropped the feature term.

What program or web site was used to tweet?
Publisher Number of tweets

Note that this table is created from the tweets () that included this information.

How many different programs were used by a user?
Number of programs Number of tweets

I have not (yet?) looked into the "multi-program" cases to see if it is people with multiple devices - perhaps a smart phone during the conference and a browser/desktop application back in the hotel room - or something else.

How chatty were people?

Below I show the distribution of the number of times an account tweeted (this excludes retweets made by the account). The number of tweets used in this analysis is and the graph accounts for of this total.

This plot shows the number of retweets by a user; as can be seen most people made 1 retweet, although there is a significant fraction () who made none. There are users that have been excluded by the cut off at ; the maximum number of retweets by a user is .

How popular is retweeting?

Below I show the distribution of the number of times a tweet was re-tweeted. The number of tweets used in this analysis is .

How long did it take for retweets to occur?

Here we look at the distribution of "retweet times" - i.e. the time between the original tweet and when it was re-tweeted. There are two graphs; the first is limited to one day, and so excludes of the distribution. The second graph shows all the tweets, but uses a much-larger bin size. The number of retweets used in this analysis is .

Would people be interested in seeing the same distribution but for replies to tweets (i.e. conversations)?

This is not a popularity contest

I am interested in seeing whether we can split up the population, so I wonder if the distribution of number of followers vs followed for each person may tell us something. Given the Twitter etiquette of reciprocity, I expected there to be a trend along the y=x line, here shown as the diagonal line. I have excluded the accounts who either have no follows or followers. Since the number of follows and followers for an account varies with time, I have used the maximum value for each account reported by the Twitter search.

The soft limit of 2000 followers can be seen as the upper bound on most of the users; I have added a blue rectangle showing the area where the number of follows is more than 2000 but the number of followers is less than 2000 to highlight this area. The limit is explained in the Twitter documentation on Are there additional limits if you are following 2000+ accounts?.

Circle area is proportional to the number of

Credits

This visualization was created using the d3.js JavaScript library.