Search

This is the preliminary ranking of WMT24 General MT systems based on automatic metrics. The official ranking will be a human evaluation, which is superior to the automatic ranking and supersedes it. The purpose of this report is not to interpret any findings but only provide preliminary results to the participants of the General MT task that may be useful during the writing of the system submission.

Annually, research teams spend large amounts of money to evaluate the quality of machine translation systems (WMT, inter alia). This is expensive because it requires a lot of expert human labor. The recently adopted annotation protocol, Error Span Annotation (ESA), has annotators marking erroneous parts of the translation and then assigning a final score. A lot of the annotator time is spent on scanning the translation for possible errors. In our work, we help the annotators by pre-filling the error annotations with recall-oriented automatic quality estimation. With this AI assistance, we obtain annotations at the same quality level while cutting down the time per span annotation by half (71s/error span $\rightarrow$ 31s/error span). The biggest advantage of ESA$^\mathrm{AI}$ protocol is an accurate priming of annotators (pre-filled error spans) before they assign the final score. This also alleviates a potential automation bias, which we confirm to be low. In addition, the annotation budget can be reduced by almost 25\% with filtering of examples that the AI deems to be very likely to be correct.

High-quality Machine Translation (MT) evaluation relies heavily on human judgments. Comprehensive error classification methods, such as Multidimensional Quality Metrics (MQM), are expensive as they are time-consuming and can only be done by experts, whose availability may be limited especially for low-resource languages. On the other hand, just assigning overall scores, like Direct Assessment (DA), is simpler and faster and can be done by translators of any level, but is less reliable. In this paper, we introduce Error Span Annotation (ESA), a human evaluation protocol which combines the continuous rating of DA with the high-level error severity span marking of MQM. We validate ESA by comparing it to MQM and DA for 12 MT systems and one human reference translation (English to German) from WMT23. The results show that ESA offers faster and cheaper annotations than MQM at the same quality level, without the requirement of expensive MQM experts.

Recent machine translation (MT) metrics calibrate their effectiveness by correlating with human judgement but without any insights about their behaviour across different error types. Challenge sets are used to probe specific dimensions of metric behaviour but there are very few such datasets and they either focus on a limited number of phenomena or a limited number of language pairs. We introduce ACES, a contrastive challenge set spanning 146 language pairs, aimed at discovering whether metrics can identify 68 translation accuracy errors. These phenomena range from simple alterations at the word/character level to more complex errors based on discourse and real-world knowledge. We conduct a large-scale study by benchmarking ACES on 50 metrics submitted to the WMT 2022 and 2023 metrics shared tasks. We benchmark metric performance, assess their incremental performance over successive campaigns, and measure their sensitivity to a range of linguistic phenomena. We also investigate claims that Large Language Models (LLMs) are effective as MT evaluators by evaluating on ACES. Our results demonstrate that different metric families struggle with different phenomena and that LLM-based methods fail to demonstrate reliable performance. Our analyses indicate that most metrics ignore the source sentence, tend to prefer surface-level overlap and end up incorporating properties of base models which are not always beneficial. We expand ACES to include error span annotations, denoted as SPAN-ACES and we use this dataset to evaluate span-based error metrics showing these metrics also need considerable improvement. Finally, we provide a set of recommendations for building better MT metrics, including focusing on error labels instead of scores, ensembling, designing strategies to explicitly focus on the source sentence, focusing on semantic content and choosing the right base model for representations., Comment: arXiv admin note: substantial text overlap with arXiv:2210.15615

Ten years ago a single metric, BLEU, governed progress in machine translation research. For better or worse, there is no such consensus today, and consequently it is difficult for researchers to develop and retain the kinds of heuristic intuitions about metric deltas that drove earlier research and deployment decisions. This paper investigates the "dynamic range" of a number of modern metrics in an effort to provide a collective understanding of the meaning of differences in scores both within and among metrics; in other words, we ask what point difference X in metric Y is required between two systems for humans to notice? We conduct our evaluation on a new large dataset, ToShip23, using it to discover deltas at which metrics achieve system-level differences that are meaningful to humans, which we measure by pairwise system accuracy. We additionally show that this method of establishing delta-accuracy is more stable than the standard use of statistical p-values in regards to testset size. Where data size permits, we also explore the effect of metric deltas and accuracy across finer-grained features such as translation direction, domain, and system closeness.

This paper introduces GEMBA-MQM, a GPT-based evaluation metric designed to detect translation quality errors, specifically for the quality estimation setting without the need for human reference translations. Based on the power of large language models (LLM), GEMBA-MQM employs a fixed three-shot prompting technique, querying the GPT-4 model to mark error quality spans. Compared to previous works, our method has language-agnostic prompts, thus avoiding the need for manual prompt preparation for new languages. While preliminary results indicate that GEMBA-MQM achieves state-of-the-art accuracy for system ranking, we advise caution when using it in academic works to demonstrate improvements over other methods due to its dependence on the proprietary, black-box GPT model., Comment: Accepted to WMT 2023

Reference-based metrics that operate at the sentence-level typically outperform quality estimation metrics, which have access only to the source and system output. This is unsurprising, since references resolve ambiguities that may be present in the source. In this paper, we investigate whether additional source context can effectively substitute for a reference. We present a metric named SLIDE (SLIding Document Evaluator), which operates on blocks of sentences. SLIDE leverages a moving window that slides over each document in the test set, feeding each chunk of sentences into an unmodified, off-the-shelf quality estimation model. We find that SLIDE obtains significantly higher pairwise system accuracy than its sentence-level baseline, in some cases even eliminating the gap with reference-base metrics. This suggests that source context may provide the same information as a human reference in disambiguating source ambiguities. This finding is especially pertinent for reference-free document-level evaluation, wherein SLIDE could provide higher-quality pairwise system assessments while only requiring document boundary annotations., Comment: NAACL 2024

Most research about natural language generation (NLG) relies on evaluation benchmarks with limited references for a sample, which may result in poor correlations with human judgements. The underlying reason is that one semantic meaning can actually be expressed in different forms, and the evaluation with a single or few references may not accurately reflect the quality of the model's hypotheses. To address this issue, this paper presents a simple and effective method, named Div-Ref, to enhance existing evaluation benchmarks by enriching the number of references. We leverage large language models (LLMs) to diversify the expression of a single reference into multiple high-quality ones to cover the semantic space of the reference sentence as much as possible. We conduct comprehensive experiments to empirically demonstrate that diversifying the expression of reference can significantly enhance the correlation between automatic evaluation and human evaluation. This idea is compatible with recent LLM-based evaluation which can similarly derive advantages from incorporating multiple references. We strongly encourage future generation benchmarks to include more references, even if they are generated by LLMs, which is once for all. We release all the code and data at https://github.com/RUCAIBox/Div-Ref to facilitate research., Comment: Accepted by NAACL 2024

Generative large language models (LLMs), e.g., ChatGPT, have demonstrated remarkable proficiency across several NLP tasks, such as machine translation, text summarization. Recent research (Kocmi and Federmann, 2023) has shown that utilizing LLMs for assessing the quality of machine translation (MT) achieves state-of-the-art performance at the system level but \textit{performs poorly at the segment level}. To further improve the performance of LLMs on MT quality assessment, we investigate several prompting designs, and propose a new prompting method called \textbf{\texttt{Error Analysis Prompting}} (EAPrompt) by combining Chain-of-Thoughts (Wei et al., 2022) and Error Analysis (Lu et al., 2023). This technique emulates the commonly accepted human evaluation framework - Multidimensional Quality Metrics (MQM, Freitag et al. (2021)) and \textit{produces explainable and reliable MT evaluations at both the system and segment level}. Experimental Results from the WMT22 metrics shared task validate the effectiveness of EAPrompt on various LLMs, with different structures. Further analysis confirms that EAPrompt effectively distinguishes major errors from minor ones, while also sharing a similar distribution of the number of errors with MQM. These findings highlight the potential of EAPrompt as a human-like evaluator prompting technique for MT evaluation., Comment: Findings of ACL 2024

We describe GEMBA, a GPT-based metric for assessment of translation quality, which works both with a reference translation and without. In our evaluation, we focus on zero-shot prompting, comparing four prompt variants in two modes, based on the availability of the reference. We investigate nine versions of GPT models, including ChatGPT and GPT-4. We show that our method for translation quality assessment only works with GPT~3.5 and larger models. Comparing to results from WMT22's Metrics shared task, our method achieves state-of-the-art accuracy in both modes when compared to MQM-based human labels. Our results are valid on the system level for all three WMT22 Metrics shared task language pairs, namely English into German, English into Russian, and Chinese into English. This provides a first glimpse into the usefulness of pre-trained, generative large language models for quality assessment of translations. We publicly release all our code and prompt templates used for the experiments described in this work, as well as all corresponding scoring results, to allow for external validation and reproducibility., Comment: Accepted in EAMT, 10 pages, 8 tables, one figure

Machine translation quality estimation (QE) predicts human judgements of a translation hypothesis without seeing the reference. State-of-the-art QE systems based on pretrained language models have been achieving remarkable correlations with human judgements yet they are computationally heavy and require human annotations, which are slow and expensive to create. To address these limitations, we define the problem of metric estimation (ME) where one predicts the automated metric scores also without the reference. We show that even without access to the reference, our model can estimate automated metrics ($\rho$=60% for BLEU, $\rho$=51% for other metrics) at the sentence-level. Because automated metrics correlate with human judgements, we can leverage the ME task for pre-training a QE model. For the QE task, we find that pre-training on TER is better ($\rho$=23%) than training for scratch ($\rho$=20%)., Comment: Accepted at EACL23 (main)

In MT evaluation, pairwise comparisons are conducted to identify the better system. In conducting the comparison, the experimenter must allocate a budget to collect Direct Assessment (DA) judgments. We provide a cost effective way to spend the budget, but show that typical budget sizes often do not allow for solid comparison. Taking the perspective that the basis of solid comparison is in achieving statistical significance, we study the power (rate of achieving significance) on a large collection of pairwise DA comparisons. Due to the nature of statistical estimation, power is low for differentiating less than 1-2 DA points, and to achieve a notable increase in power requires at least 2-3x more samples. Applying variance reduction alone will not yield these gains, so we must face the reality of undetectable differences and spending increases. In this context, we propose interim testing, an "early stopping" collection procedure that yields more power per judgment collected, which adaptively focuses the budget on pairs that are borderline significant. Interim testing can achieve up to a 27% efficiency gain when spending 3x the current budget, or 18% savings at the current evaluation power., Comment: WMT 2022

Our book "The Reality of Multi-Lingual Machine Translation" discusses the benefits and perils of using more than two languages in machine translation systems. While focused on the particular task of sequence-to-sequence processing and multi-task learning, the book targets somewhat beyond the area of natural language processing. Machine translation is for us a prime example of deep learning applications where human skills and learning capabilities are taken as a benchmark that many try to match and surpass. We document that some of the gains observed in multi-lingual translation may result from simpler effects than the assumed cross-lingual transfer of knowledge. In the first, rather general part, the book will lead you through the motivation for multi-linguality, the versatility of deep neural networks especially in sequence-to-sequence tasks to complications of this learning. We conclude the general part with warnings against too optimistic and unjustified explanations of the gains that neural networks demonstrate. In the second part, we fully delve into multi-lingual models, with a particularly careful examination of transfer learning as one of the more straightforward approaches utilizing additional languages. The recent multi-lingual techniques, including massive models, are surveyed and practical aspects of deploying systems for many languages are discussed. The conclusion highlights the open problem of machine understanding and reminds of two ethical aspects of building large-scale models: the inclusivity of research and its ecological trace., Comment: ISBN 978-80-88132-11-0. arXiv admin note: substantial text overlap with arXiv:2001.01622

Automatic metrics are commonly used as the exclusive tool for declaring the superiority of one machine translation system's quality over another. The community choice of automatic metric guides research directions and industrial developments by deciding which models are deemed better. Evaluating metrics correlations with sets of human judgements has been limited by the size of these sets. In this paper, we corroborate how reliable metrics are in contrast to human judgements on -- to the best of our knowledge -- the largest collection of judgements reported in the literature. Arguably, pairwise rankings of two systems are the most common evaluation tasks in research or deployment scenarios. Taking human judgement as a gold standard, we investigate which metrics have the highest accuracy in predicting translation quality rankings for such system pairs. Furthermore, we evaluate the performance of various metrics across different language pairs and domains. Lastly, we show that the sole use of BLEU impeded the development of improved models leading to bad deployment decisions. We release the collection of 2.3M sentence-level human judgements for 4380 systems for further analysis and replication of our work., Comment: Accepted to WMT 2021 research papers

Recent studies emphasize the need of document context in human evaluation of machine translations, but little research has been done on the impact of user interfaces on annotator productivity and the reliability of assessments. In this work, we compare human assessment data from the last two WMT evaluation campaigns collected via two different methods for document-level evaluation. Our analysis shows that a document-centric approach to evaluation where the annotator is presented with the entire document context on a screen leads to higher quality segment and document level assessments. It improves the correlation between segment and document scores and increases inter-annotator agreement for document scores but is considerably more time consuming for annotators., Comment: Presented at HumEval, EACL 2021

We present the first version of a system for interactive generation of theatre play scripts. The system is based on a vanilla GPT-2 model with several adjustments, targeting specific issues we encountered in practice. We also list other issues we encountered but plan to only solve in a future version of the system. The presented system was used to generate a theatre play script planned for premiere in February 2021., Comment: Submitted to Text2Story workshop 2021

This paper presents a description of CUNI systems submitted to the WMT20 task on unsupervised and very low-resource supervised machine translation between German and Upper Sorbian. We experimented with training on synthetic data and pre-training on a related language pair. In the fully unsupervised scenario, we achieved 25.5 and 23.7 BLEU translating from and into Upper Sorbian, respectively. Our low-resource systems relied on transfer learning from German-Czech parallel data and achieved 57.4 BLEU and 56.1 BLEU, which is an improvement of 10 BLEU points over the baseline trained only on the available small German-Upper Sorbian parallel corpus., Comment: WMT20

Gender bias in machine translation can manifest when choosing gender inflections based on spurious gender correlations. For example, always translating doctors as men and nurses as women. This can be particularly harmful as models become more popular and deployed within commercial systems. Our work presents the largest evidence for the phenomenon in more than 19 systems submitted to the WMT over four diverse target languages: Czech, German, Polish, and Russian. To achieve this, we use WinoMT, a recent automatic test suite which examines gender coreference and bias when translating from English to languages with grammatical gender. We extend WinoMT to handle two new languages tested in WMT: Polish and Czech. We find that all systems consistently use spurious correlations in the data rather than meaningful contextual information., Comment: Accepted WMT20

We present a new release of the Czech-English parallel corpus CzEng 2.0 consisting of over 2 billion words (2 "gigawords") in each language. The corpus contains document-level information and is filtered with several techniques to lower the amount of noise. In addition to the data in the previous version of CzEng, it contains new authentic and also high-quality synthetic parallel data. CzEng is freely available for research and educational purposes.

We present THEaiTRE, a starting project aimed at automatic generation of theatre play scripts. This paper reviews related work and drafts an approach we intend to follow. We plan to adopt generative neural language models and hierarchical generation approaches, supported by summarization and machine translation methods, and complemented with a human-in-the-loop approach., Comment: accepted to AI4Narratives2020

Neural machine translation is known to require large numbers of parallel training sentences, which generally prevent it from excelling on low-resource language pairs. This thesis explores the use of cross-lingual transfer learning on neural networks as a way of solving the problem with the lack of resources. We propose several transfer learning approaches to reuse a model pretrained on a high-resource language pair. We pay particular attention to the simplicity of the techniques. We study two scenarios: (a) when we reuse the high-resource model without any prior modifications to its training process and (b) when we can prepare the first-stage high-resource model for transfer learning in advance. For the former scenario, we present a proof-of-concept method by reusing a model trained by other researchers. In the latter scenario, we present a method which reaches even larger improvements in translation performance. Apart from proposed techniques, we focus on an in-depth analysis of transfer learning techniques and try to shed some light on transfer learning improvements. We show how our techniques address specific problems of low-resource languages and are suitable even in high-resource transfer learning. We evaluate the potential drawbacks and behavior by studying transfer learning in various situations, for example, under artificially damaged training corpora, or with fixed various model parts., Comment: Defended PhD thesis

Recent progress in neural machine translation is directed towards larger neural networks trained on an increasing amount of hardware resources. As a result, NMT models are costly to train, both financially, due to the electricity and hardware cost, and environmentally, due to the carbon footprint. It is especially true in transfer learning for its additional cost of training the "parent" model before transferring knowledge and training the desired "child" model. In this paper, we propose a simple method of re-using an already trained model for different language pairs where there is no need for modifications in model architecture. Our approach does not need a separate parent model for each investigated language pair, as it is typical in NMT transfer learning. To show the applicability of our method, we recycle a Transformer model trained by different researchers and use it to seed models for different language pairs. We achieve better translation quality and shorter convergence times than when training from random initialization., Comment: Accepted to EAMT 2020

Transfer learning has been proven as an effective technique for neural machine translation under low-resource conditions. Existing methods require a common target language, language relatedness, or specific training tricks and regimes. We present a simple transfer learning method, where we first train a "parent" model for a high-resource language pair and then continue the training on a lowresource pair only by replacing the training corpus. This "child" model performs significantly better than the baseline trained for lowresource pair only. We are the first to show this for targeting different languages, and we observe the improvements even for unrelated languages with different alphabets., Comment: Accepted to WMT18 reseach paper, Proceedings of the 3rd Conference on Machine Translation 2018

Skip-gram (word2vec) is a recent method for creating vector representations of words ("distributed word representations") using a neural network. The representation gained popularity in various areas of natural language processing, because it seems to capture syntactic and semantic information about words without any explicit supervision in this respect. We propose SubGram, a refinement of the Skip-gram model to consider also the word structure during the training process, achieving large gains on the Skip-gram original test set., Comment: Published at TSD 2016

Word embeddings are the interface between the world of discrete units of text processing and the continuous, differentiable world of neural networks. In this work, we examine various random and pretrained initialization methods for embeddings used in deep networks and their effect on the performance on four NLP tasks with both recurrent and convolutional architectures. We confirm that pretrained embeddings are a little better than random initialization, especially considering the speed of learning. On the other hand, we do not see any significant difference between various methods of random initialization, as long as the variance is kept reasonably low. High-variance initialization prevents the network to use the space of embeddings and forces it to use other free parameters to accomplish the task. We support this hypothesis by observing the performance in learning lexical relations and by the fact that the network can learn to perform reasonably in its task even with fixed random embeddings., Comment: Accepted to ICON 2017

We examine the effects of particular orderings of sentence pairs on the on-line training of neural machine translation (NMT). We focus on two types of such orderings: (1) ensuring that each minibatch contains sentences similar in some aspect and (2) gradual inclusion of some sentence types as the training progresses (so called "curriculum learning"). In our English-to-Czech experiments, the internal homogeneity of minibatches has no effect on the training but some of our "curricula" achieve a small improvement over the baseline., Comment: Accepted to RANLP 2017

In language identification, a common first step in natural language processing, we want to automatically determine the language of some input text. Monolingual language identification assumes that the given document is written in one language. In multilingual language identification, the document is usually in two or three languages and we just want their names. We aim one step further and propose a method for textual language identification where languages can change arbitrarily and the goal is to identify the spans of each of the languages. Our method is based on Bidirectional Recurrent Neural Networks and it performs well in monolingual and multilingual language identification tasks on six datasets covering 131 languages. The method keeps the accuracy also for short documents and across domains, so it is ideal for off-the-shelf use without preparation of training data., Comment: Accepted to EACL 2017