sentiment viz
Tweet Sentiment Visualization

This application requires IE 9, Firefox 3.6, Chrome 17.0, or Safari 5.0

This application requires IE 9, Firefox 3.6, Chrome 17.0, or Safari 5.0

This application requires IE 9, Firefox 3.6, Chrome 17.0, or Safari 5.0

xmlns="http://www.w3.org/2000/svg" width="900" height="400">

Logout


What Do I Do?
Type a keyword into the input field, then click the Query button. Recent tweets that contain your keyword are pulled from Twitter and visualized in the Sentiment tab as circles. Hover your mouse over a tweet or click on it to see its text. Words highlighted in bold blue italics or bold orange italics are the words being used to estimate the sentiment of a tweet. Blue words are evaluted as-is. Orange words are evaluated as though they are negated, for example, "happy" versus "not happy".

What Am I Seeing?
Tweets are visualized in different ways in each of the tabs at the top of the window.

Zooming
To zoom in on the tweets in the Sentiment and Topic tabs, click the zoom icon to the right of the Query button. This displays a zoom lens that you can move around the visualization.

Keywords
You can query multiple keywords at once, and combine keywords in different ways. For example, to search for tweets with the words "cat" and "dog", enter: cat dog. To search for tweets with the words "cat" or "dog", enter: cat OR dog. To search for tweets with the phrase "cat dog", enter: "cat dog". To search for tweets with the words "cat" but not "dog", enter: cat -dog.

How Do You Estimate Sentiment?
We use a sentiment dictionary to estimate sentiment. We search each tweet for words in the dictionary, then combine the words' pleasure and arousal ratings to estimate sentiment for the entire tweet. When you hover your mouse over a tweet's circle to see its text, the words in our dictionary are shown in bold italics. You can click on a tweet's circle to bring up a dialog that gives even more information.
To learn more details about how our estimates work, take a look at the tweet visualization web page.

v0.5b
Contact Author