Home

Welcome to the TweetVisualization wiki!

Project done by Wei Dai and Francois Malinowski for the Network Science class at UC Santa Barbara.

The purpose of this project is to create a tool allowing users to easily search for hashtags and view their connections through a visually appealing interface. We can see how different ideas are interconnected on Twitter, the strength of each topic as well as the strength of the connections between ideas. To accomplish this goal, Java, Twitter4J, MySQL, and d3js were used.

Essentially, this tool can be run in both online and offline modes. The offline mode provides an interface for using stored tweets extracted previously, which can be imported via a SQL database. On the other hand, the online implementation makes use of the Twitter REST API in order to pull tweets from the Internet. Both of these implementations share some graph generation methods, and the same front-end. However, due to constraints such as graph size, and resource limitations, there are some key differences as well.

The offline mode was run with a database holding around 230,000 tweets, which is a great deal more than we were able to handle with the REST API, since queries in that method are limited. As such, in order to improve response times, the entire graph is constructed upon server startup, and then queries are made through that pre-constructed graph for the purpose of obtaining subgraphs, quickly.

The online mode pulls data live from Twitter, and as such does not construct as large of a graph due to query limitations (roughly 180 requests per 15 minutes). This does not impact performance too greatly, as if we were to perform wider searches with more requests, d3js and the frontend would not be able to handle the sheet number of nodes and edges without performance deterioration.

Two videos of about 9 minutes are available on youtube to see both the online and offline modes.
Twitter hashtag network visualization (online mode)
Twitter hashtag network visualization (offline mode)