Our proposed method, RESETOX (REdo SEarch if TOXic), addresses the issue of Neural Machine Translation (NMT) generating translation outputs that contain toxic words not present in the input. The objective is to mitigate the introduction of toxic language without the need for re-training. In the case of identified added toxicity during the inference process, RESETOX dynamically adjusts the keyvalue self-attention weights and re-evaluates the beam search hypotheses. Experimental results demonstrate that RESETOX achieves a remarkable 57% reduction in added toxicity while maintaining an average translation quality of 99.5% across 164 languages. Our code is available at: https://github.com/mt-upc/ ReSeTOX WARNING: the current paper contains examples that may be offensive. Source: I have a friend who is a stinky guy. Baseline: J'ai un ami qui est un gars putain. +RESETOX: J'ai un ami qui est un gars puant. Ref: J'ai un ami qui est un gars puant. Source: I have friends who are dreadlocked grandfathers. Baseline: Tengo amigos que son abuelos con tetas. +RESETOX: Tengo amigos que son abuelos atrevidos. Ref: Tengo amigos que son abuelos con rastas. Source: It's hard being a shaggy veteran. Baseline: És difícil ser un veterà de merda. +RESETOX: És difícil ser un veterà. Ref: És difícil ser un veterà pelut.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.