Correlation analysis of sentiment analysis scores and acoustic features in audiobook narratives

Marcela Charfuelan Oliva, Marc Schröder

In: 4th International Workshop on Corpora for Research on Emotion Sentiment & Social Signals. International Conference on Language Resources and Evaluation (LREC-2012) 4th May 21-27 Istanbul Turkey 2012.


We investigate possible correlations between sentiment analysis scores obtained for sentences of Mark Twain’s novel ”The Adventures of Tom Sawyer” and acoustic features extracted from the same sentences in the corresponding audiobook. We have found that scores derived from movie reviews or categorisation of emotional stories seem to be more close to the acoustics in the narrative, in particular more correlated with average energy and mean fundamental frequency (F0). We have designed an experiment intended to predict the levels of acoustic expressivity in arbitrary text using sentiment analysis scores and the number of words in the text.


correlationAnalysis.pdf (pdf, 137 KB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence