Nodebox Experiments

Recently I have been looking into new and interesting ways of visualising data. I stumbled across an excellent suite of open source tools called nodebox (http://nodebox.net) so decided to have a play.

I have experimenting with the gui version (nodebox 3) for a couple of days and found it really quick and easy to make some nice generative looped animations. You can watch them at: https://www.instagram.com/mrlukesanger/

Wha initially interested me in nodebox is the fact it can accept CSV files. So the next step is to load up some datasets and seeing what kind of data visualisations are possible.

 

Tweet Analyser

The beta of my web app for performing sentiment analysis on tweets is online:

http://bit.ly/tweetanalyser

Simple to use. Just type in a word you want to search, press the button and wait for the result. The app collects the most recent 1000 tweets containing your keyword.

The Word Cloud tab displays the most commonly featured words that appear in tweets alongside the search keyword.

The Sentiment tab displays a table showing the overall sentiment of the tweets. The sentiment analysis algorithm is built upon the NRC Word-Emotion Association Lexicon (aka EmoLex)

 

This app was created using R and  Shiny and is intended for educational purposes only.

# ggplot2 bar chart design adapted from an excellent tutorial by J.Silge (http://juliasilge.com/blog/Joy-to-the-World/)
# word cloud adapted from shiny word cloud tutorial (http://shiny.rstudio.com/gallery/word-cloud.html)
# tweet data cleansing adapted from sentiment package tutorial (https://sites.google.com/site/miningtwitter/questions/sentiment/sentiment)
# sentiment analysis process adapted from syuzhet tutorial (https://cran.r-project.org/web/packages/syuzhet/vignettes/syuzhet-vignette.html)

Chilcot Inquiry – Sentiment Data Visualisation

Using pretty lines to visualise an ugly topic.

A short case study analysing twitter sentiment regarding the Chilcot Inquiry, a lengthy paper investigating the legalities of the UK’s involvement in the Iraq war. I collected a snapshot of 1000 tweets each day for a week after the paper was released (7/7/16 – 13/7/16) and plotted the results using R with ggplot2 package.

It is clear from the visualisation that tweets containing words associated with ‘negative’ and ‘fear’ are scoring consistently highest, with a large spike 48 hours after the report was released. This would correlate with the time taken for full media coverage on the topic.

Also worth noting is the constant low appearances of tweets containing words associated with ‘surprise’ and ‘joy’, which suggests the results of the inquiry were as predicted and brought little comfort.

The sentiment analysis was performed using my own sentiment analysis tool. You can try it out here for free: http://bit.ly/tweetanalyser