Gender bias amplification during Speed-Quality optimization in Neural Machine Translation

By Adithya Renduchintala, Denise Díaz, Kenneth Heafield, Xian Li, Mona T. Diab. 2021

42 citations recorded by the source corpus.

Is bias amplified when neural machine translation (NMT) models are optimized for speed and evaluated on generic test sets using BLEU? We investigate architectures and techniques commonly used to speed up decoding in Transformer-based models, such as greedy search, quantization, average attention networks (AANs) and shallow decoder models and show their effect on gendered noun translation. We construct a new gender bias test set, SimpleGEN, based on gendered noun phrases in which there is a single, unambiguous, correct answer. While we find minimal overall BLEU degradation as we apply speed optimizations, we observe that gendered noun translation performance degrades at a much faster rate.

Subfields: Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Safety, Robustness & Interpretability

Citations in the corpus

“I’m sorry to hear that”: Finding New Biases in Language Models with a Holistic Descriptor Dataset (2022, methodology): “…many works focus on generative models (Dinan et al., 2020a,b; Xu et al., 2021b; Kirk et al., 2021; Sheng et al., 2021b; Nozza et al., 2021; Renduchintala et al., 2021; Baheti et al., 2021; Perez et al., 2022), which are well known to pose unique challenges for automatic evaluation (Lowe…”
Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models (2021, background): “, 2021), neural machine translation (Cho et al., 2019; Stanovsky et al., 2019; Prates et al., 2020; Renduchintala et al., 2021; Savoldi et al., 2021), or language modeling (Nangia et al.”
Toxicity in Multilingual Machine Translation at Scale (2022, methodology): “Other strategies for creating bias datasets include careful handcrafting of grammars (Renduchintala et al., 2021), collecting prompts from the beginnings of existing text sentences (Dhamala et al., 2021), and swapping demographic terms in existing text, either heuristically (Ma et al., 2021; Wang…”
A Survey on Gender Bias in Natural Language Processing (2021, methodology): “WinoMT has become widely applied as a challenge dataset for gender bias detection in MT systems [Basta et al. 2020; Renduchintala et al. 2021; Saunders and Byrne 2020; Stafanovičs et al. 2020] with Saunders et al. [2020] developing a version of the WinoMT dataset with binary templates filled with…”
"I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset (2022, methodology): “…works have focused squarely on generative models (Perez et al., 2022; Xu et al., 2021b; Kirk et al., 2021a; Sheng et al., 2021b; Nozza et al., 2021; Renduchintala et al., 2021; Dinan et al., 2020a,b), which are well known to pose unique challenges for automatic evaluation (Lowe et al., 2017;…”
Square One Bias in NLP: Towards a Multi-Dimensional Exploration of the Research Manifold (2022, background): “Renduchintala et al. (2021) observe that techniques to make inference more efficient, e.g., greedy search, quantization, or shallow decoder models, have a small impact on performance, but dramatically amplify gender bias.”