Abstract:
The following problem was addressed by our project: How can we easily visualize the content of a body of text without manually analyzing its content? The initial goal was to be able to visualize captioned college lectures, but ended up being any section of text.
Our client, Mr. James Barker of Aptigent, supplied us with a large collection of captioned news reports for us to create visualizations. These television news reports were a good examples for us, since they can usually be summarized with just a few key words and relationships between words. This obviously makes them optimal for visualizing.
There were a few specifics about our task for this project. We had the ability to use a clustering program which would take a given body of text and generate, among other things, a list of keywords, which we called 'concepts,' and a list of tags. The concepts were words that the clustering program believed to have more importance, while the tags were generally words or phrases that were tied directly to one or more concepts.
Our solution needed to be web-based. In order to best accomplish this task, we chose to design our solution using HTML and JavaScript. We choose to use Raphael, a JavaScript library, to draw the visualizations.
Our solution puts a heavy emphasis on the proximity between each concept and tag. Whether or not the two appear in the same sentence is also taken into consideration.
