Reliable human evaluation is critical to the development of successful natural language generation models, but achieving it is notoriously difficult. Stability is a crucial requirement when ranking systems by quality: consistent ranking of systems across repeated evaluations is not just desirable, but essential. Without it, there is no reliable foundation for hill-climbing or product launch decisions. In this paper, we use machine translation and its state-of-the-art human evaluation framework, MQM, as a case study to understand how to set up reliable human evaluations that yield stable conclusions. We investigate the optimal configurations for item allocation to raters, number of ratings per item, and score normalization. Our study on two language pairs provides concrete recommendations for designing replicable human evaluation studies. We also collect and release the largest publicly available dataset of multi-segment translations rated by multiple professional translators, consisting of nearly 140,000 segment annotations across two language pairs.

Document-level neural machine translation (DocNMT) aims to generate translations that are both coherent and cohesive, in contrast to its sentence-level counterpart. However, due to its longer input length and limited availability of training data, DocNMT often faces the challenge of data sparsity. To overcome this issue, we propose a novel Importance-Aware Data Augmentation (IADA) algorithm for DocNMT that augments the training data based on token importance information estimated by the norm of hidden states and training gradients. We conduct comprehensive experiments on three widely-used DocNMT benchmarks. Our empirical results show that our proposed IADA outperforms strong DocNMT baselines as well as several data augmentation approaches, with statistical significance on both sentence-level and document-level BLEU.

We conduct a large-scale fine-grained comparative analysis of machine translations (MTs) against human translations (HTs) through the lens of morphosyntactic divergence. Across three language pairs and two types of divergence defined as the structural difference between the source and the target, MT is consistently more conservative than HT, with less morphosyntactic diversity, more convergent patterns, and more one-to-one alignments. Through analysis on different decoding algorithms, we attribute this discrepancy to the use of beam search that biases MT towards more convergent patterns. This bias is most amplified when the convergent pattern appears around 50% of the time in training data. Lastly, we show that for a majority of morphosyntactic divergences, their presence in HT is correlated with decreased MT performance, presenting a greater challenge for MT systems.

Existing work in document-level neural machine translation commonly concatenates several consecutive sentences as a pseudo-document, and then learns inter-sentential dependencies. This strategy limits the model’s ability to leverage information from distant context. We overcome this limitation with a novel Document Flattening (DocFlat) technique that integrates Flat-Batch Attention (FBA) and Neural Context Gate (NCG) into Transformer model to utilizes information beyond the pseudo-document boundaries. FBA allows the model to attend to all the positions in the batch and model the relationships between positions explicitly and NCG identifies the useful information from the distant context. We conduct comprehensive experiments and analyses on three benchmark datasets for English-German translation, and validate the effectiveness of two variants of DocFlat. Empirical results show that our approach outperforms strong baselines with statistical significance on BLEU, COMET and accuracy on the contrastive test set. The analyses highlight that DocFlat is highly effective in capturing the long-range information.

This paper presents the results of the WMT23 Metrics Shared Task. Participants submitting automatic MT evaluation metrics were asked to score the outputs of the translation systems competing in the WMT23 News Translation Task. All metrics were evaluated on how well they correlate with human ratings at the system and segment level. Similar to last year, we acquired our own human ratings based on expert-based human evaluation via Multidimensional Quality Metrics (MQM). Following last year’s success, we also included a challenge set subtask, where participants had to create contrastive test suites for evaluating metrics’ ability to capture and penalise specific types of translation errors. Furthermore, we improved our meta-evaluation procedure by considering fewer tasks and calculating a global score by weighted averaging across the various tasks. We present an extensive analysis on how well metrics perform on three language pairs: Chinese-English, Hebrew-English on the sentence-level and English-German on the paragraph-level. The results strongly confirm the results reported last year, that neural-based metrics are significantly better than non-neural metrics in their levels of correlation with human judgments. Further, we investigate the impact of bad reference translations on the correlations of metrics with human judgment. We present a novel approach for generating synthetic reference translations based on the collection of MT system outputs and their corresponding MQM ratings, which has the potential to mitigate bad reference issues we observed this year for some language pairs. Finally, we also study the connections between the magnitude of metric differences and their expected significance in human evaluation, which should help the community to better understand and adopt new metrics.

This paper presents the overview of the second Word-Level autocompletion (WLAC) shared task for computer-aided translation, which aims to automatically complete a target word given a translation context including a human typed character sequence. We largely adhere to the settings of the previous round of the shared task, but with two main differences: 1) The typed character sequence is obtained from the typing process of human translators to demonstrate system performance under real-world scenarios when preparing some type of testing examples; 2) We conduct a thorough analysis on the results of the submitted systems from three perspectives. From the experimental results, we observe that translation tasks are helpful to improve the performance of WLAC models. Additionally, our further analysis shows that the semantic error accounts for a significant portion of all errors, and thus it would be promising to take this type of errors into account in future.

Large, multilingual language models exhibit surprisingly good zero- or few-shot machine translation capabilities, despite having never seen the intentionally-included translation examples provided to typical neural translation systems. We investigate the role of incidental bilingualism—the unintentional consumption of bilingual signals, including translation examples—in explaining the translation capabilities of large language models, taking the Pathways Language Model (PaLM) as a case study. We introduce a mixed-method approach to measure and understand incidental bilingualism at scale. We show that PaLM is exposed to over 30 million translation pairs across at least 44 languages. Furthermore, the amount of incidental bilingual content is highly correlated with the amount of monolingual in-language content for non-English languages. We relate incidental bilingual content to zero-shot prompts and show that it can be used to mine new prompts to improve PaLM’s out-of-English zero-shot translation quality. Finally, in a series of small-scale ablations, we show that its presence has a substantial impact on translation capabilities, although this impact diminishes with model scale.

Large language models (LLMs) that have been trained on multilingual but not parallel text exhibit a remarkable ability to translate between languages. We probe this ability in an in-depth study of the pathways language model (PaLM), which has demonstrated the strongest machine translation (MT) performance among similarly-trained LLMs to date. We investigate various strategies for choosing translation examples for few-shot prompting, concluding that example quality is the most important factor. Using optimized prompts, we revisit previous assessments of PaLM’s MT capabilities with more recent test sets, modern MT metrics, and human evaluation, and find that its performance, while impressive, still lags that of state-of-the-art supervised systems. We conclude by providing an analysis of PaLM’s MT output which reveals some interesting properties and prospects for future work.

Kendall’s tau is frequently used to meta-evaluate how well machine translation (MT) evaluation metrics score individual translations. Its focus on pairwise score comparisons is intuitive but raises the question of how ties should be handled, a gray area that has motivated different variants in the literature. We demonstrate that, in settings like modern MT meta-evaluation, existing variants have weaknesses arising from their handling of ties, and in some situations can even be gamed. We propose instead to meta-evaluate metrics with a version of pairwise accuracy that gives metrics credit for correctly predicting ties, in combination with a tie calibration procedure that automatically introduces ties into metric scores, enabling fair comparison between metrics that do and do not predict ties. We argue and provide experimental evidence that these modifications lead to fairer ranking-based assessments of metric performance.

Improvements in text generation technologies such as machine translation have necessitated more costly and time-consuming human evaluation procedures to ensure an accurate signal. We investigate a simple way to reduce cost by reducing the number of text segments that must be annotated in order to accurately predict a score for a complete test set. Using a sampling approach, we demonstrate that information from document membership and automatic metrics can help improve estimates compared to a pure random sampling baseline. We achieve gains of up to 20% in average absolute error by leveraging stratified sampling and control variates. Our techniques can improve estimates made from a fixed annotation budget, are easy to implement, and can be applied to any problem with structure similar to the one we study.

Machine translation (MT) evaluation often focuses on accuracy and fluency, without paying much attention to translation style. This means that, even when considered accurate and fluent, MT output can still sound less natural than high quality human translations or text originally written in the target language. Machine translation output notably exhibits lower lexical diversity, and employs constructs that mirror those in the source sentence. In this work we propose a method for training MT systems to achieve a more natural style, i.e. mirroring the style of text originally written in the target language. Our method tags parallel training data according to the naturalness of the target side by contrasting language models trained on natural and translated data. Tagging data allows us to put greater emphasis on target sentences originally written in the target language. Automatic metrics show that the resulting models achieve lexical richness on par with human translations, mimicking a style much closer to sentences originally written in the target language. Furthermore, we find that their output is preferred by human experts when compared to the baseline translations.

This paper presents the results of the WMT22 Metrics Shared Task. Participants submitting automatic MT evaluation metrics were asked to score the outputs of the translation systems competing in the WMT22 News Translation Task on four different domains: news, social, ecommerce, and chat. All metrics were evaluated on how well they correlate with human ratings at the system and segment level. Similar to last year, we acquired our own human ratings based on expert-based human evaluation via Multidimensional Quality Metrics (MQM). This setup had several advantages, among other things: (i) expert-based evaluation is more reliable, (ii) we extended the pool of translations by 5 additional translations based on MBR decoding or rescoring which are challenging for current metrics. In addition, we initiated a challenge set subtask, where participants had to create contrastive test suites for evaluating metrics’ ability to capture and penalise specific types of translation errors. Finally, we present an extensive analysis on how well metrics perform on three language pairs: English to German, English to Russian and Chinese to English. The results demonstrate the superiority of neural-based learned metrics and demonstrate again that overlap metrics like Bleu, spBleu or chrf correlate poorly with human ratings. The results also reveal that neural-based metrics are remarkably robust across different domains and challenges.

Recent years have witnessed rapid advancements in machine translation, but the state-of-the-art machine translation system still can not satisfy the high requirements in some rigorous translation scenarios. Computer-aided translation (CAT) provides a promising solution to yield a high-quality translation with a guarantee. Unfortunately, due to the lack of popular benchmarks, the research on CAT is not well developed compared with machine translation. In this year, we hold a new shared task called Word-level AutoCompletion (WLAC) for CAT in WMT. Specifically, we introduce some resources to train a WLAC model, and particularly we collect data from CAT systems as a part of test data for this shared task. In addition, we employ both automatic and human evaluations to measure the performance of the submitted systems, and our final evaluation results reveal some findings for the WLAC task.

Reference-free evaluation has the potential to make machine translation evaluation substantially more scalable, allowing us to pivot easily to new languages or domains. It has been recently shown that the probabilities given by a large, multilingual model can achieve state of the art results when used as a reference-free metric. We experiment with various modifications to this model, and demonstrate that by scaling it up we can match the performance of BLEU. We analyze various potential weaknesses of the approach, and find that it is surprisingly robust and likely to offer reasonable performance across a broad spectrum of domains and different system qualities.

Human evaluation of modern high-quality machine translation systems is a difficult problem, and there is increasing evidence that inadequate evaluation procedures can lead to erroneous conclusions. While there has been considerable research on human evaluation, the field still lacks a commonly accepted standard procedure. As a step toward this goal, we propose an evaluation methodology grounded in explicit error analysis, based on the Multidimensional Quality Metrics (MQM) framework. We carry out the largest MQM research study to date, scoring the outputs of top systems from the WMT 2020 shared task in two language pairs using annotations provided by professional translators with access to full document context. We analyze the resulting data extensively, finding among other results a substantially different ranking of evaluated systems from the one established by the WMT crowd workers, exhibiting a clear preference for human over machine output. Surprisingly, we also find that automatic metrics based on pre-trained embeddings can outperform human crowd workers. We make our corpus publicly available for further research.

This paper presents the results of the WMT21 Metrics Shared Task. Participants were asked to score the outputs of the translation systems competing in the WMT21 News Translation Task with automatic metrics on two different domains: news and TED talks. All metrics were evaluated on how well they correlate at the system- and segment-level with human ratings. Contrary to previous years’ editions, this year we acquired our own human ratings based on expert-based human evaluation via Multidimensional Quality Metrics (MQM). This setup had several advantages: (i) expert-based evaluation has been shown to be more reliable, (ii) we were able to evaluate all metrics on two different domains using translations of the same MT systems, (iii) we added 5 additional translations coming from the same system during system development. In addition, we designed three challenge sets that evaluate the robustness of all automatic metrics. We present an extensive analysis on how well metrics perform on three language pairs: English to German, English to Russian and Chinese to English. We further show the impact of different reference translations on reference-based metrics and compare our expert-based MQM annotation with the DA scores acquired by WMT.

Conditional masked language model (CMLM) training has proven successful for non-autoregressive and semi-autoregressive sequence generation tasks, such as machine translation. Given a trained CMLM, however, it is not clear what the best inference strategy is. We formulate masked inference as a factorization of conditional probabilities of partial sequences, show that this does not harm performance, and investigate a number of simple heuristics motivated by this perspective. We identify a thresholding strategy that has advantages over the standard “mask-predict” algorithm, and provide analyses of its behavior on machine translation tasks.

There has been great progress in improving streaming machine translation, a simultaneous paradigm where the system appends to a growing hypothesis as more source content becomes available. We study a related problem in which revisions to the hypothesis beyond strictly appending words are permitted. This is suitable for applications such as live captioning an audio feed. In this setting, we compare custom streaming approaches to re-translation, a straightforward strategy where each new source token triggers a distinct translation from scratch. We find re-translation to be as good or better than state-of-the-art streaming systems, even when operating under constraints that allow very few revisions. We attribute much of this success to a previously proposed data-augmentation technique that adds prefix-pairs to the training data, which alongside wait-k inference forms a strong baseline for streaming translation. We also highlight re-translation’s ability to wrap arbitrarily powerful MT systems with an experiment showing large improvements from an upgrade to its base model.

Automatic evaluation comparing candidate translations to human-generated paraphrases of reference translations has recently been proposed by freitag2020bleu. When used in place of original references, the paraphrased versions produce metric scores that correlate better with human judgment. This effect holds for a variety of different automatic metrics, and tends to favor natural formulations over more literal (translationese) ones. In this paper we compare the results of performing end-to-end system development using standard and paraphrased references. With state-of-the-art English-German NMT components, we show that tuning to paraphrased references produces a system that is ignificantly better according to human judgment, but 5 BLEU points worse when tested on standard references. Our work confirms the finding that paraphrased references yield metric scores that correlate better with human judgment, and demonstrates for the first time that using these scores for system development can lead to significant improvements.

We consider the problem of making efficient use of heterogeneous training data in neural machine translation (NMT). Specifically, given a training dataset with a sentence-level feature such as noise, we seek an optimal curriculum, or order for presenting examples to the system during training. Our curriculum framework allows examples to appear an arbitrary number of times, and thus generalizes data weighting, filtering, and fine-tuning schemes. Rather than relying on prior knowledge to design a curriculum, we use reinforcement learning to learn one automatically, jointly with the NMT system, in the course of a single training run. We show that this approach can beat uniform baselines on Paracrawl and WMT English-to-French datasets by +3.4 and +1.3 BLEU respectively. Additionally, we match the performance of strong filtering baselines and hand-designed, state-of-the-art curricula.

Translating characters instead of words or word-fragments has the potential to simplify the processing pipeline for neural machine translation (NMT), and improve results by eliminating hyper-parameters and manual feature engineering. However, it results in longer sequences in which each symbol contains less information, creating both modeling and computational challenges. In this paper, we show that the modeling problem can be solved by standard sequence-to-sequence architectures of sufficient depth, and that deep models operating at the character level outperform identical models operating over word fragments. This result implies that alternative architectures for handling character input are better viewed as methods for reducing computation time than as improved ways of modeling longer sequences. From this perspective, we evaluate several techniques for character-level NMT, verify that they do not match the performance of our deep character baseline model, and evaluate the performance versus computation time tradeoffs they offer. Within this framework, we also perform the first evaluation for NMT of conditional computation over time, in which the model learns which timesteps can be skipped, rather than having them be dictated by a fixed schedule specified before training begins.

The past year has witnessed rapid advances in sequence-to-sequence (seq2seq) modeling for Machine Translation (MT). The classic RNN-based approaches to MT were first out-performed by the convolutional seq2seq model, which was then out-performed by the more recent Transformer model. Each of these new approaches consists of a fundamental architecture accompanied by a set of modeling and training techniques that are in principle applicable to other seq2seq architectures. In this paper, we tease apart the new architectures and their accompanying techniques in two ways. First, we identify several key modeling and training techniques, and apply them to the RNN architecture, yielding a new RNMT+ model that outperforms all of the three fundamental architectures on the benchmark WMT’14 English to French and English to German tasks. Second, we analyze the properties of each fundamental seq2seq architecture and devise new hybrid architectures intended to combine their strengths. Our hybrid models obtain further improvements, outperforming the RNMT+ model on both benchmark datasets.

In this paper, we propose a new domain adaptation technique for neural machine translation called cost weighting, which is appropriate for adaptation scenarios in which a small in-domain data set and a large general-domain data set are available. Cost weighting incorporates a domain classifier into the neural machine translation training algorithm, using features derived from the encoder representation in order to distinguish in-domain from out-of-domain data. Classifier probabilities are used to weight sentences according to their domain similarity when updating the parameters of the neural translation model. We compare cost weighting to two traditional domain adaptation techniques developed for statistical machine translation: data selection and sub-corpus weighting. Experiments on two large-data tasks show that both the traditional techniques and our novel proposal lead to significant gains, with cost weighting outperforming the traditional methods.

Neural machine translation represents an exciting leap forward in translation quality. But what longstanding weaknesses does it resolve, and which remain? We address these questions with a challenge set approach to translation evaluation and error analysis. A challenge set consists of a small set of sentences, each hand-designed to probe a system’s capacity to bridge a particular structural divergence between languages. To exemplify this approach, we present an English-French challenge set, and use it to analyze phrase-based and neural systems. The resulting analysis provides not only a more fine-grained picture of the strengths of neural systems, but also insight into which linguistic phenomena remain out of reach.

In this paper, we propose a new data selection method which uses semi-supervised convolutional neural networks based on bitokens (Bi-SSCNNs) for training machine translation systems from a large bilingual corpus. In earlier work, we devised a data selection method based on semi-supervised convolutional neural networks (SSCNNs). The new method, Bi-SSCNN, is based on bitokens, which use bilingual information. When the new methods are tested on two translation tasks (Chinese-to-English and Arabic-to-English), they significantly outperform the other three data selection methods in the experiments. We also show that the BiSSCNN method is much more effective than other methods in preventing noisy sentence pairs from being chosen for training. More interestingly, this method only needs a tiny amount of in-domain data to train the selection model, which makes fine-grained topic-dependent translation adaptation possible. In the follow-up experiments, we find that neural machine translation (NMT) is more sensitive to noisy data than statistical machine translation (SMT). Therefore, Bi-SSCNN which can effectively screen out noisy sentence pairs, can benefit NMT much more than SMT.We observed a BLEU improvement over 3 points on an English-to-French WMT task when Bi-SSCNNs were used.

Recently, there has been interest in automatically generated word classes for improving statistical machine translation (SMT) quality: e.g, (Wuebker et al, 2013). We create new models by replacing words with word classes in features applied during decoding; we call these “coarse models”. We find that coarse versions of the bilingual language models (biLMs) of (Niehues et al, 2011) yield larger BLEU gains than the original biLMs. BiLMs provide phrase-based systems with rich contextual information from the source sentence; because they have a large number of types, they suffer from data sparsity. Niehues et al (2011) mitigated this problem by replacing source or target words with parts of speech (POSs). We vary their approach in two ways: by clustering words on the source or target side over a range of granularities (word clustering), and by clustering the bilingual units that make up biLMs (bitoken clustering). We find that loglinear combinations of the resulting coarse biLMs with each other and with coarse LMs (LMs based on word classes) yield even higher scores than single coarse models. When we add an appealing “generic” coarse configuration chosen on English > French devtest data to four language pairs (keeping the structure fixed, but providing language-pair-specific models for each pair), BLEU gains on blind test data against strong baselines averaged over 5 runs are +0.80 for English > French, +0.35 for French > English, +1.0 for Arabic > English, and +0.6 for Chinese > English.

In this paper, we propose two extensions to the vector space model (VSM) adaptation technique (Chen et al., 2013b) for statistical machine translation (SMT), both of which result in significant improvements. We also systematically compare the VSM techniques to three mixture model adaptation techniques: linear mixture, log-linear mixture (Foster and Kuhn, 2007), and provenance features (Chiang et al., 2011). Experiments on NIST Chinese-to-English and Arabic-to-English tasks show that all methods achieve significant improvement over a competitive non-adaptive baseline. Except for the original VSM adaptation method, all methods yield improvements in the +1.7-2.0 BLEU range. Combining them gives further significant improvements of up to +2.6-3.3 BLEU over the baseline.

When parallel or comparable corpora are harvested from the web, there is typically a tradeoff between the size and quality of the data. In order to improve quality, corpus collection efforts often attempt to fix or remove misaligned sentence pairs. But, at the same time, Statistical Machine Translation (SMT) systems are widely assumed to be relatively robust to sentence alignment errors. However, there is little empirical evidence to support and characterize this robustness. This contribution investigates the impact of sentence alignment errors on a typical phrase-based SMT system. We confirm that SMT systems are highly tolerant to noise, and that performance only degrades seriously at very high noise levels. Our findings suggest that when collecting larger, noisy parallel data for training phrase-based SMT, cleaning up by trying to detect and remove incorrect alignments can actually degrade performance. Although fixing errors, when applicable, is a preferable strategy to removal, its benefits only become apparent for fairly high misalignment rates. We provide several explanations to support these findings.

In statistical machine translation systems, phrases with similar meanings often have similar but not identical distributions of translations. This paper proposes a new soft clustering method to smooth the conditional translation probabilities for a given phrase with those of semantically similar phrases. We call this semantic smoothing (SS). Moreover, we fabricate new phrase pairs that were not observed in training data, but which may be used for decoding. In learning curve experiments against a strong baseline, we obtain a consistent pattern of modest improvement from semantic smoothing, and further modest improvement from phrase pair fabrication.

Machine Translation traditionally treats documents as sets of independent sentences. In many genres, however, documents are highly structured, and their structure contains information that can be used to improve translation quality. We present a preliminary approach to document translation that uses structural features to modify the behaviour of a language model, at sentence-level granularity. To our knowledge, this is the first attempt to incorporate structural information into statistical MT. In experiments on structured English/French documents from the Hansard corpus, we demonstrate small but statistically significant improvements.

Cet article décrit une approche combinant différents modèles statistiques pour la traduction automatique basée sur les segments. Pour ce faire, différentes ressources sont utilisées, dont deux corpus parallèles aux caractéristiques différentes et un dictionnaire de terminologie bilingue et ce, afin d’améliorer la performance quantitative et qualitative du système de traduction. Nous évaluons notre approche sur la paire de langues français-anglais et montrons comment la combinaison des ressources proposées améliore de façon significative les résultats.

We describe an experiment in rapid development of a statistical machine translation (SMT) system from scratch, using limited resources: under this heading we include not only training data, but also computing power, linguistic knowledge, programming effort, and absolute time.

Text prediction is a form of interactive machine translation that is well suited to skilled translators. In recent work it has been shown that simple statistical translation models can be applied within a usermodeling framework to improve translator productivity by over 10% in simulated results. For the sake of efficiency in making real-time predictions, these models ignore the alignment relation between source and target texts. In this paper we introduce a new model that captures fuzzy alignments in a very simple way, and show that it gives modest improvements in predictive performance without significantly increasing the time required to generate predictions.

In this paper, we present a way to integrate bilingual lexicons into an operational probabilistic translation assistant (TransType). These lexicons could be any resource available to the translator (e.g. terminological lexicons) or any resource statistically derived from training material. We describe a bilingual lexicon acquisition process that we developped and we evaluate from a theoretical point of view its benefits to a translation completion task.