Despite its growth as a focus within natural language processing, sentiment analysis is often limited to the examination of a single dimension of sentiment—polarity—which measures the relative positivity, neutrality, or negativity of the language of a text. The analysis presented here combines three additional sentiment dimensions—aspect, mood, and intensity—into a new sentiment metric. This novel metric provides a single sentiment score that includes all four dimensions and is more comprehensive than a polarity score alone. The usefulness of the new metric is demonstrated first by applying it to complaints filed with the U.S. Consumer Financial Protection Bureau and correlating the scores with the outcomes of the cases. The analysis demonstrates that consumers received better outcomes when sentiments expressed in their complaints had a more positive comprehensive score. Next, the new metric is applied to tweets sent by former U.S. president Donald Trump. The scores are shown to distinguish the tweets authored by President Trump from tweets authored by others that the president retweeted. The correlation between the comprehensive sentiment scores of these two types of tweets demonstrates that President Trump retweeted others’ messages that were more negative in their sentiment expression than those he authored himself.

Multidimensional sentiment analysis consumer review analysis social media authorship Trump tweets comprehensive sentiment metric cosent sentiment metric