The advantages of lexicon-based sentiment analysis in an age of machine learning.

We demonstrate the strong performance of lexicon-based sentiment analysis using MultiLexScaled, an approach which averages valences across a number of widely-used general-purpose lexica. We validate it against benchmark datasets from a range of different domains, comparing performance against machine learning and LLM alternatives. In addition, we illustrate the value of identifying fine-grained sentiment levels by showing, in an analysis of pre- and post- 9/11 British press coverage of Muslims, that binarized valence metrics give rise to different (and erroneous) conclusions about the nature of the post-9/11 shock as well as about differences between broadsheet and tabloid coverage.

STAIR students over the years have been instrumental in helping test and develop the python notebooks we have put on Github to allow others to easily use the method. Check out Github to see more.

Covering Muslims

We present the first systematic, large-scale analysis of American newspaper coverage of Muslims. By comparing it over time with reporting on other groups and issues as well as coverage of the subject in other countries, we demonstrate conclusively how negative American newspapers have been in their treatment of Muslims across the two-decade period between 1996 and 2016, both in an absolute sense and compared to a range of other groups. The same pattern holds in other countries, such as Australia, Canada, and the UK. While 9/11 did not make coverage more negative in the long run, it did dramatically increase the prevalence of references to terrorism and extremism.