Sentiment expression conditioned by
affective transitions and social forces

Moritz Sudhof, Andrés Gómez Emilsson, Andrew L. Maas, and Christopher Potts

Emotional states are not independent, random, or isolated. They proceed along systematic paths and are conditioned by the states of others in our communities. We develop a simple probability model for capturing the conditional dependencies between emotional states. We use CRFs to incorporate information about emotion sequences and show that this delivers improvements over models that classify text independently of context.

Read the Paper Check out our Poster Get the Slides

emotion-transition-map

Bibtex

@inproceedings{Sudhof-etal:2014,
     Author = {Sudhof, Moritz and G{\'o}mez Emilsson, Andr{\'e}s and Maas, Andrew L. and Potts, Christopher},
     Month = {August},
     Title = {Sentiment Expression Conditioned by Affective Transitions and Social Forces},
     Booktitle = {Proceedings of 20th Conference on Knowledge Discovery and Data Mining},
     Pages = {1136--1145},
     Location = {New York},
     Publisher = {ACM},
     Address = {New York},
     Year = {2014},
     Doi = {10.1145/2623330.2623687}}