Investigation of Gender Bias in Turkish Word Embeddings
Nurullah Sevim, Aykut Koç · 2021
Investigating gender bias in Natural Language Processing has recently gained importance due to the negative consequences of a possible sexist approach. Especially by examining such biases in English word embeddings in various contexts, many studies have been conducted on these issues. In this study, the status of Turkish word embeddings in terms of gender bias was examined and the Turkish language structure was compared with English within the scope of gender biases. As a result of the measurement of gender bias in word embeddings, it was concluded that Turkish contains less gender bias in language structure compared to English.