Wide range screening of algorithmic bias in word embedding models using large sentiment lexicons reveals underreported bias types
Fig 2
To validate the methodology used in subsequent analysis, this figure shows that systematically projecting a large sentiment lexicon (HGI) on cultural axes with widespread agreed-upon positive/negative polarity results in positive terms in the lexicon being associated with the poles representing life, health, democracy and respectful historical figures.
Conversely, negative terms tend to be associated with the poles representing death, disease, dictatorship and malevolent historical figures.