Twitter Search, Assange and Impure

Julian Assange and WikiLeaks are still in the news this week. After what is arguably the most public security breach in history, Assange has become something of a pariah, with companies such as Amazon, Paypal, EveryDNS (not EasyDNS), the Australian Government and credit cards cutting services.

Issues like the politics of freedom of discourse, the internet and who gets to decide where journalism stops and espionage begins have been hotly debated with varying degrees of sophistication all over the internet. Discussion of the discussion has not been as widespread.

Visualising Twitter Search

Using Twitter search data, visualisations and natural language tools for categorising words by frequency in language and occurrences in the text, it is possible to take a snapshot of discussion topics and sentiment.

Impure makes it easy to generate these visualisations quickly and easily for any other trending topic that may develop around a brand. The main limitation is in the amount of data that Impure inports from Twitter. Collating the information elsewhere and importing it as a table is one way of dealing with this restriction.

The first issue with the data set was the words ‘Julian’ and ‘Assange’. As the Twitter search was on the term ‘Assange’ and both words were rather remarkable, occurred often and were uncommon natural language, they skewed the graph heavily. As they can be assumed, the terms were removed from the text string, and using a logarithmic scale made the visualisation more useable.

Twitter Timeline & User Activity

ObjectsOnTime makes it possible to display the tweets sequentially. Adding repetitions by user as a weighting to the visualisation shows which tweets were made by those discussing the subject often, those who were not and what was actually said. Unsuprisingly the most active user in this Twitter search dataset was a bot.

The data used by the workspace will change each time it is accessed. As only a small sample is created each time the work space is accessed there can be a lot of variance in activity. This is an interesting demonstration of natural language tools and timelines to see who is saying what and demonstrating novel behaviours online.

Explore the Workspace