Skip to content
cdschillaci edited this page Mar 4, 2015 · 9 revisions

Mood Prediction

Use the LiveJournal moodid tag as ground truth for predicting mood from blog text.

<current_moodid> tags

There are 132 mood ID numbers ranging from 1 to 134. These are listed in a csv file. Note that there are no tags 50, 94.

Background in understanding affect

Russell, "A circumplex model of affect" (1980). http://dx.doi.org/10.1037/h0077714
The version we are using for our first pass: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2367156/

Previous Work in Machine Learning

Keshtkar and Inkpen, "A hierarchical approach to mood classification in blogs" (2011). http://dx.doi.org/10.1017/S1351324911000118
An analysis of LiveJournal data using the tree structure from LiveJournal to improve accuracy of classifying to all of the 132 moods. Good summary statistics on the data are displayed as well. Do we have the same data set?