For our assignment, Bianca Datta and I created a simple web page that demonstrates a possible tool for journalists in understanding a lot of social media data. We took the idea of creating a story via social media and attempted to automate it a bit more.
On the left is a feed of tweets based on the topic ‘St. Patrick’s Day’. The search terms we used were ‘StPatty’, ‘StPaddy’, and ‘StPatrick’ . Once we had these tweets in our database, we processed them to get topics using LDA via a Node.js package. The top ranking topics were then displaying on the right side of the webpage using a very simple web graphic. Based on the term’s use, the topic would be allotted a larger height for more use and smaller height for less use. The left column shows common topic terms in tweets with the word ‘Patty’. The middle column shows topic terms for ‘Paddy’ and the right column shows topic terms for ‘Patrick’.
This is a screenshot of the page:
We also gathered content from Instagram, Vine, and Flickr and processed each of those to get topic rankings.
link : http://web.media.mit.edu/~vdiep/vdiepbdatta/indexy.html
Vivian & Bianca, i liked the idea of using LDA on curated tweets to get topics. I think, the richer the content, the better the results for the LDA algorithm. Do you find the number of topics algorithmically or are you specifying the topics in the algorithm. The link (http://web.media.mit.edu/~vdiep/vdiepbdatta/indexy.html) is not functional. It does not load the content properly. Would love to see an interactive demo of this.