Extract, analyze and visualize Twitter data - no code needed.

By Dr. Walid Al-Saqaf, data scientist & journalism researcher

Although there are many tools out there that researchers and journalists could use to analyze Twitter data, many require technical know-hows while others cost money.

Recognizing the deficiency in this area, Mecodify is a tool created to help fill this gap. Initially, its main aim was to provide researchers in the Media, Conflict and Democratization (MeCoDEM) project with a way to easily extract, analyze and visualize Twitter data without the hassle of coding. The aim was to build a powerful solution from the ground-up that would be easy to use, free of charge and open source. It should be possible to install and run on a regular computer or on a server. The tool should be able to extract data from up to half a million tweets for a particular case, and have multiple cases that can be analyzed independently. The data generated by the tool had to be exportable to other applications and any dynamic graphs produced had to be embeddable.

While still under development, Mecodify has largely met those requirements and is going through further improvements. Since it was officially launched in Brussels in November 2016, it has been used to generate findings that are being used to write academic articles. Some researchers and journalists outside MeCoDEM have also found it of use.

Mecodify’s deployment does require a bit of effort but, once installed, it becomes a powerful tool that researchers and journalists can use to dig into the world of Twitter in ways that were otherwise not possible.

Investigating the tweets

Take the example of a case around the February 2016 general elections in Uganda. The case was created by using the hashtags #UgandaElections and #UgandaDecides, which resulted in over 60,000 tweets posted between 13 and 20 February 2016. By default, Mecodify plots the level of tweets and retweets on the timeline but also allows overlaying (stacking) graphs to represent the number of unique users who tweeted over the same period. Additionally, it can plot the number of original tweets on top of the other two graphs to give a fuller picture of the level of activity that happened during the given period, as shown below.


It is important to note is that the graph is interactive, and allows you to zoom in and click on any of the points to get the full list of tweets that were posted at that particular time. In addition to the many filters that can be adjusted to focus on one subsection of the tweets (for example, only images, tweets from verified accounts, replies, and more), it is also possible to explore the internal interactions between users in the given corpus. For example, the below snapshot is taken from a conversation that took place on Twitter between three individual accounts.


Such interactions are possible to trace recursively, making the tool quite effective to follow discussions by tracking which of the tweets is in response to which. Additionally, Mecodify allows you to filter tweets that are in response to a particular tweet. This can be useful when visualizing the level of interaction over time. The types and level of filtering that Mecodify offers are numerous and span the type of tweet, the language, and even the software used to post that tweet (also called source), as shown in the below figure.


Investigating the tweeters

In addition to revealing insights about the content of tweets, Mecodify allows you to investigate the users that produced these tweets, also called ‘tweeters’. In a separate tab, the tool identifies the most prolific tweeters, the most followed, the most mentioned, and the most replied to. The figure below shows the most followed accounts that had tweets in the Ugandan corpus.


It is also possible to visualize the network of mentions and responses in an interactive form. By clicking on a node in this graph, other nodes that a reply or sent are revealed. The figure below shows that the account @nbstv received replies as well as sent out replies to other accounts.


Finally, Mecodify allows you to export this user data, including data related to mentions and replies, in CSV format. The connections data are possible to import in another third party platform called Kumu, which helps to organize this complex information into interactive relationship maps. An example of an interactive network visualized by Kumu is shown in the figure below.


Kumu allows for clustering, filtering and the execution of a diverse set of social network analysis metrics such as degree centrality, eigenvector, reach, and more. This makes it easy for a Mecodify user conduct an analysis of both the content of tweets as well as the networks that are formed by the tweeters of this content.

As a new tool that helps study and discover many complex structures and messages on Twitter, Mecodify is already being used by several researchers in the social sciences. However, it is data journalists and scientists who would find it even more useful, particularly as Twitter has become a major platform for generating new stories from around the globe.

Visit the Mecodify webpage here.