Sentiment and structure in word co-occurrence networks on Twitter

We explore the relationship between context and happiness scores in political tweets using word co-occurrence networks, where nodes network are words, weight of an edge is number corpus for which two connected words co-occur. In particular, we consider with hashtags #imwithher #crookedhillary, both relating to Hillary Clinton's presidential bid 2016. then analyze properties conjunction by comparing null models separate effects structure score distribution. Neutral found be dominant most regardless polarity, tend co-occur neutral words. do not observe any homophily among positive negative However, when perform backboning, community detection results groupings meaningful narratives, each group correspond its respective theme. Thus, although no clear at node or level, a community-centric approach can isolate themes competing sentiments corpus.