From newspaper to microblogging: What does it take to find opinions?
Sidorenko Wladimir, Sonntag Jonathan, Krüger Nina, Stieglitz Stefan, Stede Manfred
We investigate the differences and the levelsof difficulty for sentiment analysis on thetwo genres of newspaper text and twitter text(tweets). Two existing systems are comparedwith respect to their performance on bothgenres: SentiStrength (Thelwall et al., 2012)and SO-CAL (Taboada et al., 2011). Bothhave similar architectures, using hand-builtpolarity dictionaries and rules for combiningsentiment values in context. SentiStrength,however, has been geared specifically towardshort social-media text, whereas SO-CAL wasbuilt for general, longer text. After the initialcomparison, we successively enrich theSO-CAL-based analysis with tweet-specificmechanisms and observe that in some cases,this improves the performance. A qualitativeerror analysis then identifies classes of typicalproblems the two systems have with tweets.
Sentiment, Social Media, Analytics