Explainable NLP techniques primarily explain by answering "Which tokens in the input are responsible for this prediction?''. We argue that for NLP models that make predictions by comparing two input texts, it is more useful to explain by answering "What differences between the two inputs explain this prediction?''. We introduce a technique to generate contrastive highlights that explain the predictions of a semantic divergence model via phrase-alignment-guided erasure. We show that the resulting highlights match human rationales of cross-lingual semantic differences better than popular post-hoc saliency techniques and that they successfully help people detect fine-grained meaning differences in human translations and critical machine translation errors.
Translations help people understand content written in another language. However, even correct literal translations do not fulfill that goal when people lack the necessary background to understand them. Professional translators incorporate explicitations to explain the missing context by considering cultural differences between source and target audiences. Despite its potential to help users, NLP research on explicitation is limited because of the dearth of adequate evaluation methods. This work introduces techniques for automatically generating explicitations, motivated by WikiExpl: a dataset that we collect from Wikipedia and annotate with human translators. The resulting explicitations are useful as they help answer questions more accurately in a multilingual question answering framework.
Lexical ambiguity is a challenging and pervasive problem in machine translation (\mt). We introduce a simple and scalable approach to resolve translation ambiguity by incorporating a small amount of extra-sentential context in neural \mt. Our approach requires no sense annotation and no change to standard model architectures. Since actual document context is not available for the vast majority of \mt training data, we collect related sentences for each input to construct pseudo-documents. Salient words from pseudo-documents are then encoded as a prefix to each source sentence to condition the generation of the translation. To evaluate, we release \docmucow, a challenge set for translation disambiguation based on the English-German \mucow \cite{raganato-etal-2020-evaluation} augmented with document IDs. Extensive experiments show that our method translates ambiguous source words better than strong sentence-level baselines and comparable document-level baselines while reducing training costs.
Despite the progress we have recorded in scaling multilingual machine translation (MT) models and evaluation data to several under-resourced African languages, it is difficult to measure accurately the progress we have made on these languages because evaluation is often performed on n-gram matching metrics like BLEU that often have worse correlation with human judgments. Embedding-based metrics such as COMET correlate better; however, lack of evaluation data with human ratings for under-resourced languages, complexity of annotation guidelines like Multidimensional Quality Metrics (MQM), and limited language coverage of multilingual encoders have hampered their applicability to African languages. In this paper, we address these challenges by creating high-quality human evaluation data with a simplified MQM guideline for error-span annotation and direct assessment (DA) scoring for 13 typologically diverse African languages. Furthermore, we develop AfriCOMET, a COMET evaluation metric for African languages by leveraging DA training data from high-resource languages and African-centric multilingual encoder (AfroXLM-Roberta) to create the state-of-the-art evaluation metric for African languages MT with respect to Spearman-rank correlation with human judgments (+0.406).
A major challenge in the practical use of Machine Translation (MT) is that users lack guidance to make informed decisions about when to rely on outputs. Progress in quality estimation research provides techniques to automatically assess MT quality, but these techniques have primarily been evaluated in vitro by comparison against human judgments outside of a specific context of use. This paper evaluates quality estimation feedback in vivo with a human study simulating decision-making in high-stakes medical settings. Using Emergency Department discharge instructions, we study how interventions based on quality estimation versus backtranslation assist physicians in deciding whether to show MT outputs to a patient. We find that quality estimation improves appropriate reliance on MT, but backtranslation helps physicians detect more clinically harmful errors that QE alone often misses.
Recent research at the intersection of AI explainability and fairness has focused on how explanations can improve human-plus-AI task performance as assessed by fairness measures. We propose to characterize what constitutes an explanation that is itself "fair" -- an explanation that does not adversely impact specific populations. We formulate a novel evaluation method of "fair explanations" using not just accuracy and label time, but also psychological impact of explanations on different user groups across many metrics (mental discomfort, stereotype activation, and perceived workload). We apply this method in the context of content moderation of potential hate speech, and its differential impact on Asian vs. non-Asian proxy moderators, across explanation approaches (saliency map and counterfactual explanation). We find that saliency maps generally perform better and show less evidence of disparate impact (group) and individual unfairness than counterfactual explanations. Content warning: This paper contains examples of hate speech and racially discriminatory language. The authors do not support such content. Please consider your risk of discomfort carefully before continuing reading!
Text simplification rewrites text to be more readable for a specific audience, while preserving its meaning. However, determining what makes a text easy to read depends on who are the intended readers. Recent work has introduced a wealth of techniques to control output simplicity, ranging from specifying the desired reading grade level to providing control tokens that directly encode low-level simplification edit operations. However, it remains unclear how to set the input parameters that control simplification in practice. Existing approaches set them at the corpus level, disregarding the complexity of individual source text, and do not directly evaluate them at the instance level. In this work, we conduct an empirical study to understand how different control mechanisms impact the adequacy and simplicity of model outputs. Based on these insights, we introduce a simple method for predicting control tokens at the sentence level to enhance the quality of the simplified text. Predicting control token values using features extracted from the original complex text and a user-specified degree of complexity improves the quality of the simplified outputs over corpus-level search-based heuristics.
AI systems have shown impressive performance at answering questions by retrieving relevant context. However, with the increasingly large models, it is impossible and often undesirable to constrain models' knowledge or reasoning to only the retrieved context. This leads to a mismatch between the information that these models access to derive the answer and the information available to the user consuming the AI predictions to assess the AI predicted answer. In this work, we study how users interact with AI systems in absence of sufficient information to assess AI predictions. Further, we ask the question of whether adding the requisite background alleviates the concerns around over-reliance in AI predictions. Our study reveals that users rely on AI predictions even in the absence of sufficient information needed to assess its correctness. Providing the relevant background, however, helps users catch AI errors better, reducing over-reliance on incorrect AI predictions. On the flip side, background information also increases users' confidence in their correct as well as incorrect judgments. Contrary to common expectation, aiding a user's perusal of the context and the background through highlights is not helpful in alleviating the issue of over-confidence stemming from availability of more information. Our work aims to highlight the gap between how NLP developers perceive informational need in human-AI interaction and the actual human interaction with the information available to them.
Neural sequence generation models are known to "hallucinate", by producing outputs that are unrelated to the source text. These hallucinations are potentially harmful, yet it remains unclear in what conditions they arise and how to mitigate their impact. In this work, we first identify internal model symptoms of hallucinations by analyzing the relative token contributions to the generation in contrastive hallucinated vs. non-hallucinated outputs generated via source perturbations. We then show that these symptoms are reliable indicators of natural hallucinations, by using them to design a lightweight hallucination detector which outperforms both model-free baselines and strong classifiers based on quality estimation or large pre-trained models on manually annotated English-Chinese and German-English translation test beds.
Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License.