PLOS One publication: The advantages of lexicon-based sentiment analysis in an age of machine learning.
The sentiment analysis method used throughout Covering Muslims to assess the negativity of Muslim coverage has at last found a published home at the journal PLOS One.
By Maurits van der Veen in publications
November 1, 2024
A core feature of Covering Muslims is the analysis of how positive/negative newspaper coverage of Muslim is. Performing this analysis requires a robust, fine-grained sentiment analysis method. Since we were unable to find an existing method that proved acceptable for our purposes, we developed our own, MultiLexScaled, which uses multiple sentiment lexica, calibrated and scaled on a representative corpus of newspaper articles.
The paper introducing the method, long available in preprint form, has now found an official published home at the general science journal PLOS One. Below is one of the figures from the paper, which demonstrates the importance of fine-grained sentiment analysis, as compared to the binary approach (positive or negative) more common in machine learning approaches. Click on the figure to go to the article.
