Question answering (QA) systems are now available through numerous commercial applications for a wide variety of domains, serving millions of users that interact with them via speech interfaces. However, current benchmarks in QA research do not account for the errors that speech recognition models might introduce, nor do they consider the language variations (dialects) of the users. To address this gap, we augment an existing QA dataset to construct a multi-dialect, spoken QA benchmark on five languages (Arabic, Bengali, English, Kiswahili, Korean) with more than 68k audio prompts in 24 dialects from 255 speakers. We provide baseline results showcasing the real-world performance of QA systems and analyze the effect of language variety and other sensitive speaker attributes on downstream performance. Last, we study the fairness of the ASR and QA models with respect to the underlying user populations.

Language domains that require very careful use of terminology are abundant and reflect a significant part of the translation industry. In this work we introduce a benchmark for evaluating the quality and consistency of terminology translation, focusing on the medical (and COVID-19 specifically) domain for five language pairs: English to French, Chinese, Russian, and Korean, as well as Czech to German. We report the descriptions and results of the participating systems, commenting on the need for further research efforts towards both more adequate handling of terminologies as well as towards a proper formulation and evaluation of the task.

The performance of neural machine translation (NMT) systems only trained on a single language variant degrades when confronted with even slightly different language variations. With this work, we build upon previous work to explore how to mitigate this issue. We show that fine-tuning using naturally occurring noise along with pseudo-references (i.e. “corrected” non-native inputs translated using the baseline NMT system) is a promising solution towards systems robust to such type of input variations. We focus on four translation pairs, from English to Spanish, Italian, French, and Portuguese, with our system achieving improvements of up to 3.1 BLEU points compared to the baselines, establishing a new state-of-the-art on the JFLEG-ES dataset. All datasets and code are publicly available here: https://github.com/mahfuzibnalam/finetuning_for_robustness .