Mixwest 2013 Twitter Analytics Day Two

Mixwest

Yesterday, I attended the second day of a two-day conference here in Indianapolis called Mixwest, which focuses on four core areas: marketing, social media, design and tech. It reminds me of the ScienceOnline conference but smaller, local (Midwest), and much more focused on business.

As I did yesterday for Mixwest 2013 Day 1, I mined all of the tweets about the conference from the Twitter stream.  I’m interested in organizing unstructured data and textual analytics. Below are Twitter analytics for Mixwest 2013 Day 2.

First, some statistics

Tweets were collected and analyzed from 7 am Friday morning August 9th, 2013 (2 hours before the conference started) to 6 pm the same day (1 hour after the conference ended). In total, there were 1,670 tweets — just 43 less than day 1 — consisting of 26,835 words from 226 people. There were 241 people tweeting on day 1, so that’s 15 less tweeters. The top tweeter was Kayla Hulen (Twitter: @kayla_hulen) with 74 tweets!

Here are the top 10 Twitter users and their number of tweets:

Rank Twitter user Number of tweets
1 @kayla_hulen 74
2 @bnpositive 61
3 @mixwest 60
4 @theramennoodle
58
5 @heatherchastain 54
6 @robbyslaughter 50
7 @juettj 48
8 @melissabalkon 46
9 @socialreactor 46
10 @edeckers 43

 

Like yesterday, again I didn’t make the top 10 and came in at number 21! For day 2, I attribute the lack of Tweets to Nathan Hand (Twitter: @nathan_hand) and his session “Using social media to find Bigfoot”. In case you were wondering, we did find him and no, we’re not telling you. Thanks for an interesting, engaging session Nathan!

Similar to day 1, most conference attendees using Twitter actually tweeted very little: 155 people tweeted just 1-4 times. Twenty people tweeted 5-9 times and 25 people tweeted 10-19 times. Each of the other groups with 20 or more tweets consisted of less than 15 people.

Tweets per tweeters

Let’s take a moment and focus on retweets. In my opinion, retweets are a way of saying “yeah, I like that” or “I agree!” and for the purposes of this analysis come in two forms: they can be a simple rebroadcast (e.g. RT: “original tweet”), which we’ll call a RT without conversation, or a retweet with comments back to the original tweeter (e.g. Agreed! RT: “original tweet”), which we’ll call a RT with conversation.

Without getting into details, a modified tweet (MT) is also classified as a RT with conversation, since usually the modification is made to provide extra space to comment.

Just under one-third (31%) of the #mixwest13 tweets today were retweets. Of those, 9% (12 of 522) were a RT with conversation. The top retweeter was Mixwest (Twitter: @Mixwest) with 41 retweets.

Mixwest13 day2 retweets

#Mixwest13 Day 2 word cloud of tweets

I love word clouds. I think they’re a useful visual representation of data and effectively convey “themes” of a data set.

From 26,835 words, 3,249 were unique. I calculated the frequency of all 3,249 words and then “cleaned” the data, removing all words less than 3 characters and common words such as “your”, “from”, “that” or “about”.

Word frequencies were adjusted — that is, blunted — at the high end such that any frequencies greater than 100 were set to 100, and any frequencies greater than 150 were set to 150. This was done to prevent distortion of the word cloud.

The top 100 terms were then imported into Wordle and a weighted tag cloud was generated. Feel free to download any of the files below and reshare.

Mixwest13 Day 2 Word Cloud

Here are the top 10 terms (prior to adjustment) from the word cloud above:

Rank Term Frequency
1 mixwest13 1679
2 @hunckler 145
3 @mixwest 144
4 @douglaskarr 121
5 about 117
6 @colefarrell 117
7 @edeckers 99
8 people 90
9 @robbyslaughter 84
10 social 79

All data and images are available for download: low-resolution image, high-resolution image or raw data set of 100 terms with frequencies.

Walter Jessen is a digital strategist, writer, web developer and data scientist. You can typically find him behind the screen something with an internet connection.

  • Sally

    Hi Walter,

    Great blog post.

    I’ve recently discovered Wordle and I’m really interested in the method you used to achieve this:

    “From 26,835 words, 3,249 were unique. I calculated the frequency of
    all 3,249 words and then “cleaned” the data, removing all words less
    than 3 characters and common words such as “your”, “from”, “that” or
    “about”.

    Word frequencies were adjusted — that is, blunted — at the high end
    such that any frequencies greater than 100 were set to 100, and any
    frequencies greater than 150 were set to 150. This was done to prevent
    distortion of the word cloud.”

    I produce word clouds based on tweets from a particular audience I monitor. I export the tweets but would like to “clean” them up, similarly to how you have done. Is this something you can do in excel? I’m happy with the weighting that Wordle produces, but I would like to get rid of words such as “just” “now” “use” “please” etc which appear in the word clouds I’m producing.

    Please let me know if you need any further information.

    • http://www.walterjessen.com/ Walter Jessen

      Hi Sally,

      Start by removing words with 3 letters or less. Then sort the word list by frequency; you’ll have to manually remove the common words.

      Good luck!