Lexical methods for bias exploration from a Latin American perspective

LUCIANA BENOTTI; LAURA ALONSO ALEMANY; LUCIA GONZALEZ

Workshop; Workshop Widening Access to Natural Language Processing; 2022

This short paper briefly introduces a long termproject called EDIA. The goal of this projectis to design and test a methodology that allows social scientists and domain experts inLatin America to explore social biases presentin word embeddings, and to define the kinds ofbiases they want to explore. In particular, weinvestigate how domain experts and social scientists can intuitively manipulate lexical representations, resorting to their expertise to assessboth hypothesized and emergent biases.