Andy Way

Yasmin Moslem, Rejwanul Haque, Andy Way

Research on Machine Translation (MT) has achieved important breakthroughs in several areas. While there is much more to be done in order to build on this success, we believe that the language industry needs better ways to take full advantage of current achievements. Due to a combination of factors, including time, resources, and skills, businesses tend to apply pragmatism into their AI workflows. Hence, they concentrate more on outcomes, e.g. delivery, shipping, releases, and features, and adopt high-level working production solutions, where possible. Among the features thought to be helpful for translators are sentence-level and word-level translation auto-suggestion and auto-completion. Suggesting alternatives can inspire translators and limit their need to refer to external resources, which hopefully boosts their productivity. This work describes our submissions to WMT's shared task on word-level auto-completion, for the Chinese-to-English, English-to-Chinese, German-to-English, and English-to-German language directions. We investigate the possibility of using pre-trained models and out-of-the-box features from available libraries. We employ random sampling to generate diverse alternatives, which reveals good results. Furthermore, we introduce our open-source API, based on CTranslate2, to serve translations, auto-suggestions, and auto-completions.

Yasmin Moslem, Rejwanul Haque, John D. Kelleher et al.

Consistency is a key requirement of high-quality translation. It is especially important to adhere to pre-approved terminology and adapt to corrected translations in domain-specific projects. Machine translation (MT) has achieved significant progress in the area of domain adaptation. However, real-time adaptation remains challenging. Large-scale language models (LLMs) have recently shown interesting capabilities of in-context learning, where they learn to replicate certain input-output text generation patterns, without further fine-tuning. By feeding an LLM at inference time with a prompt that consists of a list of translation pairs, it can then simulate the domain and style characteristics. This work aims to investigate how we can utilize in-context learning to improve real-time adaptive MT. Our extensive experiments show promising results at translation time. For example, LLMs can adapt to a set of in-domain sentence pairs and/or terminology while translating a new sentence. We observe that the translation quality with few-shot in-context learning can surpass that of strong encoder-decoder MT systems, especially for high-resource languages. Moreover, we investigate whether we can combine MT from strong encoder-decoder models with fuzzy matches, which can further improve translation quality, especially for less supported languages. We conduct our experiments across five diverse language pairs, namely English-to-Arabic (EN-AR), English-to-Chinese (EN-ZH), English-to-French (EN-FR), English-to-Kinyarwanda (EN-RW), and English-to-Spanish (EN-ES).

Yasmin Moslem, Rejwanul Haque, John D. Kelleher et al.

Preservation of domain knowledge from the source to target is crucial in any translation workflow. It is common in the translation industry to receive highly specialized projects, where there is hardly any parallel in-domain data. In such scenarios where there is insufficient in-domain data to fine-tune Machine Translation (MT) models, producing translations that are consistent with the relevant context is challenging. In this work, we propose a novel approach to domain adaptation leveraging state-of-the-art pretrained language models (LMs) for domain-specific data augmentation for MT, simulating the domain characteristics of either (a) a small bilingual dataset, or (b) the monolingual source text to be translated. Combining this idea with back-translation, we can generate huge amounts of synthetic bilingual in-domain data for both use cases. For our investigation, we use the state-of-the-art Transformer architecture. We employ mixed fine-tuning to train models that significantly improve translation of in-domain texts. More specifically, in both scenarios, our proposed methods achieve improvements of approximately 5-6 BLEU and 2-3 BLEU, respectively, on the Arabic-to-English and English-to-Arabic language pairs. Furthermore, the outcome of human evaluation corroborates the automatic evaluation results.

Yasmin Moslem, Gianfranco Romani, Mahdi Molaei et al.

This paper discusses the methods that we used for our submissions to the WMT 2023 Terminology Shared Task for German-to-English (DE-EN), English-to-Czech (EN-CS), and Chinese-to-English (ZH-EN) language pairs. The task aims to advance machine translation (MT) by challenging participants to develop systems that accurately translate technical terms, ultimately enhancing communication and understanding in specialised domains. To this end, we conduct experiments that utilise large language models (LLMs) for two purposes: generating synthetic bilingual terminology-based data, and post-editing translations generated by an MT model through incorporating pre-approved terms. Our system employs a four-step process: (i) using an LLM to generate bilingual synthetic data based on the provided terminology, (ii) fine-tuning a generic encoder-decoder MT model, with a mix of the terminology-based synthetic data generated in the first step and a randomly sampled portion of the original generic training data, (iii) generating translations with the fine-tuned MT model, and (iv) finally, leveraging an LLM for terminology-constrained automatic post-editing of the translations that do not include the required terms. The results demonstrate the effectiveness of our proposed approach in improving the integration of pre-approved terms into translations. The number of terms incorporated into the translations of the blind dataset increases from an average of 36.67% with the generic model to an average of 72.88% by the end of the process. In other words, successful utilisation of terms nearly doubles across the three language pairs.

Longyue Wang, Zhaopeng Tu, Yan Gu et al.

Translating literary works has perennially stood as an elusive dream in machine translation (MT), a journey steeped in intricate challenges. To foster progress in this domain, we hold a new shared task at WMT 2023, the first edition of the Discourse-Level Literary Translation. First, we (Tencent AI Lab and China Literature Ltd.) release a copyrighted and document-level Chinese-English web novel corpus. Furthermore, we put forth an industry-endorsed criteria to guide human evaluation process. This year, we totally received 14 submissions from 7 academia and industry teams. We employ both automatic and human evaluations to measure the performance of the submitted systems. The official ranking of the systems is based on the overall human judgments. In addition, our extensive analysis reveals a series of interesting findings on literary and discourse-aware MT. We release data, system outputs, and leaderboard at http://www2.statmt.org/wmt23/literary-translation-task.html.

Inacio Vieira, Will Allred, Séamus Lankford et al.

Decoder-only LLMs have shown impressive performance in MT due to their ability to learn from extensive datasets and generate high-quality translations. However, LLMs often struggle with the nuances and style required for organisation-specific translation. In this study, we explore the effectiveness of fine-tuning Large Language Models (LLMs), particularly Llama 3 8B Instruct, leveraging translation memories (TMs), as a valuable resource to enhance accuracy and efficiency. We investigate the impact of fine-tuning the Llama 3 model using TMs from a specific organisation in the software sector. Our experiments cover five translation directions across languages of varying resource levels (English to Brazilian Portuguese, Czech, German, Finnish, and Korean). We analyse diverse sizes of training datasets (1k to 207k segments) to evaluate their influence on translation quality. We fine-tune separate models for each training set and evaluate their performance based on automatic metrics, BLEU, chrF++, TER, and COMET. Our findings reveal improvement in translation performance with larger datasets across all metrics. On average, BLEU and COMET scores increase by 13 and 25 points, respectively, on the largest training set against the baseline model. Notably, there is a performance deterioration in comparison with the baseline model when fine-tuning on only 1k and 2k examples; however, we observe a substantial improvement as the training dataset size increases. The study highlights the potential of integrating TMs with LLMs to create bespoke translation models tailored to the specific needs of businesses, thus enhancing translation quality and reducing turn-around times. This approach offers a valuable insight for organisations seeking to leverage TMs and LLMs for optimal translation outcomes, especially in narrower domains.

Steinþór Steingrímsson, Hrafn Loftsson, Andy Way

We present SentAlign, an accurate sentence alignment tool designed to handle very large parallel document pairs. Given user-defined parameters, the alignment algorithm evaluates all possible alignment paths in fairly large documents of thousands of sentences and uses a divide-and-conquer approach to align documents containing tens of thousands of sentences. The scoring function is based on LaBSE bilingual sentence representations. SentAlign outperforms five other sentence alignment tools when evaluated on two different evaluation sets, German-French and English-Icelandic, and on a downstream machine translation task.

Séamus Lankford, Haithem Afli, Andy Way

The advent of Multilingual Language Models (MLLMs) and Large Language Models has spawned innovation in many areas of natural language processing. Despite the exciting potential of this technology, its impact on developing high-quality Machine Translation (MT) outputs for low-resource languages remains relatively under-explored. Furthermore, an open-source application, dedicated to both fine-tuning MLLMs and managing the complete MT workflow for low-resources languages, remains unavailable. We aim to address these imbalances through the development of adaptMLLM, which streamlines all processes involved in the fine-tuning of MLLMs for MT. This open-source application is tailored for developers, translators, and users who are engaged in MT. An intuitive interface allows for easy customisation of hyperparameters, and the application offers a range of metrics for model evaluation and the capability to deploy models as a translation service directly within the application. As a multilingual tool, we used adaptMLLM to fine-tune models for two low-resource language pairs: English to Irish (EN$\leftrightarrow$GA) and English to Marathi (EN$\leftrightarrow$MR). Compared with baselines from the LoResMT2021 Shared Task, the adaptMLLM system demonstrated significant improvements. In the EN$\rightarrow$GA direction, an improvement of 5.2 BLEU points was observed and an increase of 40.5 BLEU points was recorded in the GA$\rightarrow$EN direction. Significant improvements in the translation performance of the EN$\leftrightarrow$MR pair were also observed notably in the MR$\rightarrow$EN direction with an increase of 21.3 BLEU points. Finally, a fine-grained human evaluation of the MLLM output on the EN$\rightarrow$GA pair was conducted using the Multidimensional Quality Metrics and Scalar Quality Metrics error taxonomies. The application and models are freely available.

Séamus Lankford, Haithem Afli, Andy Way

adaptNMT streamlines all processes involved in the development and deployment of RNN and Transformer neural translation models. As an open-source application, it is designed for both technical and non-technical users who work in the field of machine translation. Built upon the widely-adopted OpenNMT ecosystem, the application is particularly useful for new entrants to the field since the setup of the development environment and creation of train, validation and test splits is greatly simplified. Graphing, embedded within the application, illustrates the progress of model training, and SentencePiece is used for creating subword segmentation models. Hyperparameter customization is facilitated through an intuitive user interface, and a single-click model development approach has been implemented. Models developed by adaptNMT can be evaluated using a range of metrics, and deployed as a translation service within the application. To support eco-friendly research in the NLP space, a green report also flags the power consumption and kgCO$_{2}$ emissions generated during model development. The application is freely available.

Séamus Lankford, Haithem Afli, Andy Way

adaptNMT is an open-source application that offers a streamlined approach to the development and deployment of Recurrent Neural Networks and Transformer models. This application is built upon the widely-adopted OpenNMT ecosystem, and is particularly useful for new entrants to the field, as it simplifies the setup of the development environment and creation of train, validation, and test splits. The application offers a graphing feature that illustrates the progress of model training, and employs SentencePiece for creating subword segmentation models. Furthermore, the application provides an intuitive user interface that facilitates hyperparameter customization. Notably, a single-click model development approach has been implemented, and models developed by adaptNMT can be evaluated using a range of metrics. To encourage eco-friendly research, adaptNMT incorporates a green report that flags the power consumption and kgCO${_2}$ emissions generated during model development. The application is freely available.

Yasmin Moslem, Rejwanul Haque, Andy Way

This paper presents the outcomes of fine-tuning Mistral 7B, a general-purpose large language model (LLM), for adaptive machine translation (MT). The fine-tuning process involves utilising a combination of zero-shot and one-shot translation prompts within the medical domain. The primary objective is to enhance real-time adaptive MT capabilities of Mistral 7B, enabling it to adapt translations to the required domain at inference time. The results, particularly for Spanish-to-English MT, showcase the efficacy of the fine-tuned model, demonstrating quality improvements in both zero-shot and one-shot translation scenarios, surpassing Mistral 7B's baseline performance. Notably, the fine-tuned Mistral outperforms ChatGPT "gpt-3.5-turbo" in zero-shot translation while achieving comparable one-shot translation quality. Moreover, the zero-shot translation of the fine-tuned Mistral matches NLLB 3.3B's performance, and its one-shot translation quality surpasses that of NLLB 3.3B. These findings emphasise the significance of fine-tuning efficient LLMs like Mistral 7B to yield high-quality zero-shot translations comparable to task-oriented models like NLLB 3.3B. Additionally, the adaptive gains achieved in one-shot translation are comparable to those of commercial LLMs such as ChatGPT. Our experiments demonstrate that, with a relatively small dataset of 20,000 segments that incorporate a mix of zero-shot and one-shot prompts, fine-tuning significantly enhances Mistral's in-context learning ability, especially for real-time adaptive MT.

Séamus Lankford, Haithem Afli, Andy Way

The Transformer model is the state-of-the-art in Machine Translation. However, in general, neural translation models often under perform on language pairs with insufficient training data. As a consequence, relatively few experiments have been carried out using this architecture on low-resource language pairs. In this study, hyperparameter optimization of Transformer models in translating the low-resource English-Irish language pair is evaluated. We demonstrate that choosing appropriate parameters leads to considerable performance improvements. Most importantly, the correct choice of subword model is shown to be the biggest driver of translation performance. SentencePiece models using both unigram and BPE approaches were appraised. Variations on model architectures included modifying the number of layers, testing various regularisation techniques and evaluating the optimal number of heads for attention. A generic 55k DGT corpus and an in-domain 88k public admin corpus were used for evaluation. A Transformer optimized model demonstrated a BLEU score improvement of 7.8 points when compared with a baseline RNN model. Improvements were observed across a range of metrics, including TER, indicating a substantially reduced post editing effort for Transformer optimized models with 16k BPE subword models. Bench-marked against Google Translate, our translation engines demonstrated significant improvements. The question of whether or not Transformers can be used effectively in a low-resource setting of English-Irish translation has been addressed. Is féidir linn - yes we can.

Séamus Lankford, Haithem Afli, Andy Way

In this study, a human evaluation is carried out on how hyperparameter settings impact the quality of Transformer-based Neural Machine Translation (NMT) for the low-resourced English--Irish pair. SentencePiece models using both Byte Pair Encoding (BPE) and unigram approaches were appraised. Variations in model architectures included modifying the number of layers, evaluating the optimal number of heads for attention and testing various regularisation techniques. The greatest performance improvement was recorded for a Transformer-optimized model with a 16k BPE subword model. Compared with a baseline Recurrent Neural Network (RNN) model, a Transformer-optimized model demonstrated a BLEU score improvement of 7.8 points. When benchmarked against Google Translate, our translation engines demonstrated significant improvements. Furthermore, a quantitative fine-grained manual evaluation was conducted which compared the performance of machine translation systems. Using the Multidimensional Quality Metrics (MQM) error taxonomy, a human evaluation of the error types generated by an RNN-based system and a Transformer-based system was explored. Our findings show the best-performing Transformer system significantly reduces both accuracy and fluency errors when compared with an RNN-based model.

Séamus Lankford, Haithem Afli, Andy Way

Translation models for the specific domain of translating Covid data from English to Irish were developed for the LoResMT 2021 shared task. Domain adaptation techniques, using a Covid-adapted generic 55k corpus from the Directorate General of Translation, were applied. Fine-tuning, mixed fine-tuning and combined dataset approaches were compared with models trained on an extended in-domain dataset. As part of this study, an English-Irish dataset of Covid related data, from the Health and Education domains, was developed. The highest-performing model used a Transformer architecture trained with an extended in-domain Covid dataset. In the context of this study, we have demonstrated that extending an 8k in-domain baseline dataset by just 5k lines improved the BLEU score by 27 points.

Longyue Wang, Siyou Liu, Chenyang Lyu et al.

Following last year, we have continued to host the WMT translation shared task this year, the second edition of the Discourse-Level Literary Translation. We focus on three language directions: Chinese-English, Chinese-German, and Chinese-Russian, with the latter two ones newly added. This year, we totally received 10 submissions from 5 academia and industry teams. We employ both automatic and human evaluations to measure the performance of the submitted systems. The official ranking of the systems is based on the overall human judgments. We release data, system outputs, and leaderboard at https://www2.statmt.org/wmt24/literary-translation-task.html.

Séamus Lankford, Haithem Afli, Órla Ní Loinsigh et al.

Machine Translation is a mature technology for many high-resource language pairs. However in the context of low-resource languages, there is a paucity of parallel data datasets available for developing translation models. Furthermore, the development of datasets for low-resource languages often focuses on simply creating the largest possible dataset for generic translation. The benefits and development of smaller in-domain datasets can easily be overlooked. To assess the merits of using in-domain data, a dataset for the specific domain of health was developed for the low-resource English to Irish language pair. Our study outlines the process used in developing the corpus and empirically demonstrates the benefits of using an in-domain dataset for the health domain. In the context of translating health-related data, models developed using the gaHealth corpus demonstrated a maximum BLEU score improvement of 22.2 points (40%) when compared with top performing models from the LoResMT2021 Shared Task. Furthermore, we define linguistic guidelines for developing gaHealth, the first bilingual corpus of health data for the Irish language, which we hope will be of use to other creators of low-resource data sets. gaHealth is now freely available online and is ready to be explored for further research.

Joss Moorkens, Andy Way, Séamus Lankford

While the previous chapters have shown how machine translation (MT) can be useful, in this chapter we discuss some of the side-effects and risks that are associated, and how they might be mitigated. With the move to neural MT and approaches using Large Language Models (LLMs), there is an associated impact on climate change, as the models built by multinational corporations are massive. They are hugely expensive to train, consume large amounts of electricity, and output huge volumes of kgCO2 to boot. However, smaller models which still perform to a high level of quality can be built with much lower carbon footprints, and tuning pre-trained models saves on the requirement to train from scratch. We also discuss the possible detrimental effects of MT on translators and other users. The topics of copyright and ownership of data are discussed, as well as ethical considerations on data and MT use. Finally, we show how if done properly, using MT in crisis scenarios can save lives, and we provide a method of how this might be done.

Séamus Lankford, Andy Way

In an evolving landscape of crisis communication, the need for robust and adaptable Machine Translation (MT) systems is more pressing than ever, particularly for low-resource languages. This study presents a comprehensive exploration of leveraging Large Language Models (LLMs) and Multilingual LLMs (MLLMs) to enhance MT capabilities in such scenarios. By focusing on the unique challenges posed by crisis situations where speed, accuracy, and the ability to handle a wide range of languages are paramount, this research outlines a novel approach that combines the cutting-edge capabilities of LLMs with fine-tuning techniques and community-driven corpus development strategies. At the core of this study is the development and empirical evaluation of MT systems tailored for two low-resource language pairs, illustrating the process from initial model selection and fine-tuning through to deployment. Bespoke systems are developed and modelled on the recent Covid-19 pandemic. The research highlights the importance of community involvement in creating highly specialised, crisis-specific datasets and compares custom GPTs with NLLB-adapted MLLM models. It identifies fine-tuned MLLM models as offering superior performance compared with their LLM counterparts. A scalable and replicable model for rapid MT system development in crisis scenarios is outlined. Our approach enhances the field of humanitarian technology by offering a blueprint for developing multilingual communication systems during emergencies.

Liangyou Li, Andy Way, Qun Liu

We present graph-based translation models which translate source graphs into target strings. Source graphs are constructed from dependency trees with extra links so that non-syntactic phrases are connected. Inspired by phrase-based models, we first introduce a translation model which segments a graph into a sequence of disjoint subgraphs and generates a translation by combining subgraph translations left-to-right using beam search. However, similar to phrase-based models, this model is weak at phrase reordering. Therefore, we further introduce a model based on a synchronous node replacement grammar which learns recursive translation rules. We provide two implementations of the model with different restrictions so that source graphs can be parsed efficiently. Experiments on Chinese--English and German--English show that our graph-based models are significantly better than corresponding sequence- and tree-based baselines.

Alberto Poncelas, Jan Buts, James Hadley et al.

Building Machine Translation (MT) systems for low-resource languages remains challenging. For many language pairs, parallel data are not widely available, and in such cases MT models do not achieve results comparable to those seen with high-resource languages. When data are scarce, it is of paramount importance to make optimal use of the limited material available. To that end, in this paper we propose employing the same parallel sentences multiple times, only changing the way the words are split each time. For this purpose we use several Byte Pair Encoding models, with various merge operations used in their configuration. In our experiments, we use this technique to expand the available data and improve an MT system involving a low-resource language pair, namely English-Esperanto. As an additional contribution, we made available a set of English-Esperanto parallel data in the literary domain.

Alberto Poncelas, Pintu Lohar, Andy Way et al.

Sentiment classification has been crucial for many natural language processing (NLP) applications, such as the analysis of movie reviews, tweets, or customer feedback. A sufficiently large amount of data is required to build a robust sentiment classification system. However, such resources are not always available for all domains or for all languages. In this work, we propose employing a machine translation (MT) system to translate customer feedback into another language to investigate in which cases translated sentences can have a positive or negative impact on an automatic sentiment classifier. Furthermore, as performing a direct translation is not always possible, we explore the performance of automatic classifiers on sentences that have been translated using a pivot MT system. We conduct several experiments using the above approaches to analyse the performance of our proposed sentiment classification system and discuss the advantages and drawbacks of classifying translated sentences.

Xabier Soto, Dimitar Shterionov, Alberto Poncelas et al.

Machine translation (MT) has benefited from using synthetic training data originating from translating monolingual corpora, a technique known as backtranslation. Combining backtranslated data from different sources has led to better results than when using such data in isolation. In this work we analyse the impact that data translated with rule-based, phrase-based statistical and neural MT systems has on new MT systems. We use a real-world low-resource use-case (Basque-to-Spanish in the clinical domain) as well as a high-resource language pair (German-to-English) to test different scenarios with backtranslation and employ data selection to optimise the synthetic corpora. We exploit different data selection strategies in order to reduce the amount of data used, while at the same time maintaining high-quality MT systems. We further tune the data selection method by taking into account the quality of the MT systems used for backtranslation and lexical diversity of the resulting corpora. Our experiments show that incorporating backtranslated data from different sources can be beneficial, and that availing of data selection can yield improved performance.

Andy Way, Rejwanul Haque, Guodong Xie et al.

Every day, more people are becoming infected and dying from exposure to COVID-19. Some countries in Europe like Spain, France, the UK and Italy have suffered particularly badly from the virus. Others such as Germany appear to have coped extremely well. Both health professionals and the general public are keen to receive up-to-date information on the effects of the virus, as well as treatments that have proven to be effective. In cases where language is a barrier to access of pertinent information, machine translation (MT) may help people assimilate information published in different languages. Our MT systems trained on COVID-19 data are freely available for anyone to use to help translate information published in German, French, Italian, Spanish into English, as well as the reverse direction.

Alberto Poncelas, Wichaya Pidchamook, Chao-Hong Liu et al.

Thai is a low-resource language, so it is often the case that data is not available in sufficient quantities to train an Neural Machine Translation (NMT) model which perform to a high level of quality. In addition, the Thai script does not use white spaces to delimit the boundaries between words, which adds more complexity when building sequence to sequence models. In this work, we explore how to augment a set of English--Thai parallel data by replicating sentence-pairs with different word segmentation methods on Thai, as training data for NMT model training. Using different merge operations of Byte Pair Encoding, different segmentations of Thai sentences can be obtained. The experiments show that combining these datasets, performance is improved for NMT models trained with a dataset that has been split using a supervised splitting tool.

Alberto Poncelas, Mohammad Aboomar, Jan Buts et al.

Optical character recognition (OCR) for historical documents is a complex procedure subject to a unique set of material issues, including inconsistencies in typefaces and low quality scanning. Consequently, even the most sophisticated OCR engines produce errors. This paper reports on a tool built for postediting the output of Tesseract, more specifically for correcting common errors in digitized historical documents. The proposed tool suggests alternatives for word forms not found in a specified vocabulary. The assumed error is replaced by a presumably correct alternative in the post-edition based on the scores of a Language Model (LM). The tool is tested on a chapter of the book An Essay Towards Regulating the Trade and Employing the Poor of this Kingdom (Cary ,1719). As demonstrated below, the tool is successful in correcting a number of common errors. If sometimes unreliable, it is also transparent and subject to human intervention.

Georg Rehm, Katrin Marheinecke, Stefanie Hegele et al.

Multilingualism is a cultural cornerstone of Europe and firmly anchored in the European treaties including full language equality. However, language barriers impacting business, cross-lingual and cross-cultural communication are still omnipresent. Language Technologies (LTs) are a powerful means to break down these barriers. While the last decade has seen various initiatives that created a multitude of approaches and technologies tailored to Europe's specific needs, there is still an immense level of fragmentation. At the same time, AI has become an increasingly important concept in the European Information and Communication Technology area. For a few years now, AI, including many opportunities, synergies but also misconceptions, has been overshadowing every other topic. We present an overview of the European LT landscape, describing funding programmes, activities, actions and challenges in the different countries with regard to LT, including the current state of play in industry and the LT market. We present a brief overview of the main LT-related activities on the EU level in the last ten years and develop strategic guidance with regard to four key dimensions.

Alberto Poncelas, Andy Way

Neural Machine Translation (NMT) models tend to achieve best performance when larger sets of parallel sentences are provided for training. For this reason, augmenting the training set with artificially-generated sentence pairs can boost performance. Nonetheless, the performance can also be improved with a small number of sentences if they are in the same domain as the test set. Accordingly, we want to explore the use of artificially-generated sentences along with data-selection algorithms to improve German-to-English NMT models trained solely with authentic data. In this work, we show how artificially-generated sentences can be more beneficial than authentic pairs, and demonstrate their advantages when used in combination with data-selection algorithms.

Eva Vanmassenhove, Christian Hardmeier, Andy Way

Speakers of different languages must attend to and encode strikingly different aspects of the world in order to use their language correctly (Sapir, 1921; Slobin, 1996). One such difference is related to the way gender is expressed in a language. Saying "I am happy" in English, does not encode any additional knowledge of the speaker that uttered the sentence. However, many other languages do have grammatical gender systems and so such knowledge would be encoded. In order to correctly translate such a sentence into, say, French, the inherent gender information needs to be retained/recovered. The same sentence would become either "Je suis heureux", for a male speaker or "Je suis heureuse" for a female one. Apart from morphological agreement, demographic factors (gender, age, etc.) also influence our use of language in terms of word choices or even on the level of syntactic constructions (Tannen, 1991; Pennebaker et al., 2003). We integrate gender information into NMT systems. Our contribution is two-fold: (1) the compilation of large datasets with speaker information for 20 language pairs, and (2) a simple set of experiments that incorporate gender information into NMT for multiple language pairs. Our experiments show that adding a gender feature to an NMT system significantly improves the translation quality for some language pairs.

Alberto Poncelas, Maja Popovic, Dimitar Shterionov et al.

Neural Machine Translation (NMT) models achieve their best performance when large sets of parallel data are used for training. Consequently, techniques for augmenting the training set have become popular recently. One of these methods is back-translation (Sennrich et al., 2016), which consists on generating synthetic sentences by translating a set of monolingual, target-language sentences using a Machine Translation (MT) model. Generally, NMT models are used for back-translation. In this work, we analyze the performance of models when the training data is extended with synthetic data using different MT approaches. In particular we investigate back-translated data generated not only by NMT but also by Statistical Machine Translation (SMT) models and combinations of both. The results reveal that the models achieve the best performances when the training set is augmented with back-translated data created by merging different MT approaches.

Alberto Poncelas, Gideon Maillette de Buy Wenniger, Andy Way

Machine Translation models are trained to translate a variety of documents from one language into another. However, models specifically trained for a particular characteristics of the documents tend to perform better. Fine-tuning is a technique for adapting an NMT model to some domain. In this work, we want to use this technique to adapt the model to a given test set. In particular, we are using transductive data selection algorithms which take advantage the information of the test set to retrieve sentences from a larger parallel set. In cases where the model is available at translation time (when the test set is provided), it can be adapted with a small subset of data, thereby achieving better performance than a generic model or a domain-adapted model.

Eva Vanmassenhove, Dimitar Shterionov, Andy Way

This work presents an empirical approach to quantifying the loss of lexical richness in Machine Translation (MT) systems compared to Human Translation (HT). Our experiments show how current MT systems indeed fail to render the lexical diversity of human generated or translated text. The inability of MT systems to generate diverse outputs and its tendency to exacerbate already frequent patterns while ignoring less frequent ones, might be the underlying cause for, among others, the currently heavily debated issues related to gender biased output. Can we indeed, aside from biased data, talk about an algorithm that exacerbates seen biases?

Alberto Poncelas, Gideon Maillette de Buy Wenniger, Andy Way

Data selection has proven its merit for improving Neural Machine Translation (NMT), when applied to authentic data. But the benefit of using synthetic data in NMT training, produced by the popular back-translation technique, raises the question if data selection could also be useful for synthetic data? In this work we use Infrequent N-gram Recovery (INR) and Feature Decay Algorithms (FDA), two transductive data selection methods to obtain subsets of sentences from synthetic data. These methods ensure that selected sentences share n-grams with the test set so the NMT model can be adapted to translate it. Performing data selection on back-translated data creates new challenges as the source-side may contain noise originated by the model used in the back-translation. Hence, finding n-grams present in the test set become more difficult. Despite that, in our work we show that adapting a model with a selection of synthetic data is an useful approach.

Gideon Maillette de Buy Wenniger, Lambert Schomaker, Andy Way

Neural handwriting recognition (NHR) is the recognition of handwritten text with deep learning models, such as multi-dimensional long short-term memory (MDLSTM) recurrent neural networks. Models with MDLSTM layers have achieved state-of-the art results on handwritten text recognition tasks. While multi-directional MDLSTM-layers have an unbeaten ability to capture the complete context in all directions, this strength limits the possibilities for parallelization, and therefore comes at a high computational cost. In this work we develop methods to create efficient MDLSTM-based models for NHR, particularly a method aimed at eliminating computation waste that results from padding. This proposed method, called example-packing, replaces wasteful stacking of padded examples with efficient tiling in a 2-dimensional grid. For word-based NHR this yields a speed improvement of factor 6.6 over an already efficient baseline of minimal padding for each batch separately. For line-based NHR the savings are more modest, but still significant. In addition to example-packing, we propose: 1) a technique to optimize parallelization for dynamic graph definition frameworks including PyTorch, using convolutions with grouping, 2) a method for parallelization across GPUs for variable-length example batches. All our techniques are thoroughly tested on our own PyTorch re-implementation of MDLSTM-based NHR models. A thorough evaluation on the IAM dataset shows that our models are performing similar to earlier implementations of state-of-the-art models. Our efficient NHR model and some of the reusable techniques discussed with it offer ways to realize relatively efficient models for the omnipresent scenario of variable-length inputs in deep learning.

Eva Vanmassenhove, Amit Moryossef, Alberto Poncelas et al.

We present our system for the CLIN29 shared task on cross-genre gender detection for Dutch. We experimented with a multitude of neural models (CNN, RNN, LSTM, etc.), more "traditional" models (SVM, RF, LogReg, etc.), different feature sets as well as data pre-processing. The final results suggested that using tokenized, non-lowercased data works best for most of the neural models, while a combination of word clusters, character trigrams and word lists showed to be most beneficial for the majority of the more "traditional" (that is, non-neural) models, beating features used in previous tasks such as n-grams, character n-grams, part-of-speech tags and combinations thereof. In contradiction with the results described in previous comparable shared tasks, our neural models performed better than our best traditional approaches with our best feature set-up. Our final model consisted of a weighted ensemble model combining the top 25 models. Our final model won both the in-domain gender prediction task and the cross-genre challenge, achieving an average accuracy of 64.93% on the in-domain gender prediction task, and 56.26% on cross-genre gender prediction.

Alberto Poncelas, Andy Way, Kepa Sarasola

In this paper we present the ADAPT system built for the Basque to English Low Resource MT Evaluation Campaign. Basque is a low-resourced, morphologically-rich language. This poses a challenge for Neural Machine Translation models which usually achieve better performance when trained with large sets of data. Accordingly, we used synthetic data to improve the translation quality produced by a model built using only authentic data. Our proposal uses back-translated data to: (a) create new sentences, so the system can be trained with more data; and (b) translate sentences that are close to the test set, so the model can be fine-tuned to the document to be translated.

Alberto Poncelas, Gideon Maillette de Buy Wenniger, Andy Way

Data selection techniques applied to neural machine translation (NMT) aim to increase the performance of a model by retrieving a subset of sentences for use as training data. One of the possible data selection techniques are transductive learning methods, which select the data based on the test set, i.e. the document to be translated. A limitation of these methods to date is that using the source-side test set does not by itself guarantee that sentences are selected with correct translations, or translations that are suitable given the test-set domain. Some corpora, such as subtitle corpora, may contain parallel sentences with inaccurate translations caused by localization or length restrictions. In order to try to fix this problem, in this paper we propose to use an approximated target-side in addition to the source-side when selecting suitable sentence-pairs for training a model. This approximated target-side is built by pre-translating the source-side. In this work, we explore the performance of this general idea for one specific data selection approach called Feature Decay Algorithms (FDA). We train German-English NMT models on data selected by using the test set (source), the approximated target side, and a mixture of both. Our findings reveal that models built using a combination of outputs of FDA (using the test set and an approximated target side) perform better than those solely using the test set. We obtain a statistically significant improvement of more than 1.5 BLEU points over a model trained with all data, and more than 0.5 BLEU points over a strong FDA baseline that uses source-side information only.

Longyue Wang, Zhaopeng Tu, Andy Way et al.

Pronouns are frequently omitted in pro-drop languages, such as Chinese, generally leading to significant challenges with respect to the production of complete translations. Recently, Wang et al. (2018) proposed a novel reconstruction-based approach to alleviating dropped pronoun (DP) translation problems for neural machine translation models. In this work, we improve the original model from two perspectives. First, we employ a shared reconstructor to better exploit encoder and decoder representations. Second, we jointly learn to translate and predict DPs in an end-to-end manner, to avoid the errors propagated from an external DP prediction model. Experimental results show that our approach significantly improves both translation performance and DP prediction accuracy.

Jinhua Du, Jingguang Han, Andy Way et al.

Attention mechanisms are often used in deep neural networks for distantly supervised relation extraction (DS-RE) to distinguish valid from noisy instances. However, traditional 1-D vector attention models are insufficient for the learning of different contexts in the selection of valid instances to predict the relationship for an entity pair. To alleviate this issue, we propose a novel multi-level structured (2-D matrix) self-attention mechanism for DS-RE in a multi-instance learning (MIL) framework using bidirectional recurrent neural networks. In the proposed method, a structured word-level self-attention mechanism learns a 2-D matrix where each row vector represents a weight distribution for different aspects of an instance regarding two entities. Targeting the MIL issue, the structured sentence-level attention learns a 2-D matrix where each row vector represents a weight distribution on selection of different valid in-stances. Experiments conducted on two publicly available DS-RE datasets show that the proposed framework with a multi-level structured self-attention mechanism significantly outperform state-of-the-art baselines in terms of PR curves, P@N and F1 measures.

Antonio Toral, Sheila Castilho, Ke Hu et al.

We reassess a recent study (Hassan et al., 2018) that claimed that machine translation (MT) has reached human parity for the translation of news from Chinese into English, using pairwise ranking and considering three variables that were not taken into account in that previous study: the language in which the source side of the test set was originally written, the translation proficiency of the evaluators, and the provision of inter-sentential context. If we consider only original source text (i.e. not translated from another language, or translationese), then we find evidence showing that human parity has not been achieved. We compare the judgments of professional translators against those of non-experts and discover that those of the experts result in higher inter-annotator agreement and better discrimination between human and machine translations. In addition, we analyse the human translations of the test set and identify important translation issues. Finally, based on these findings, we provide a set of recommendations for future human evaluations of MT.

Peyman Passban, Qun Liu, Andy Way

Recently, neural machine translation (NMT) has emerged as a powerful alternative to conventional statistical approaches. However, its performance drops considerably in the presence of morphologically rich languages (MRLs). Neural engines usually fail to tackle the large vocabulary and high out-of-vocabulary (OOV) word rate of MRLs. Therefore, it is not suitable to exploit existing word-based models to translate this set of languages. In this paper, we propose an extension to the state-of-the-art model of Chung et al. (2016), which works at the character level and boosts the decoder with target-side morphological information. In our architecture, an additional morphology table is plugged into the model. Each time the decoder samples from a target vocabulary, the table sends auxiliary signals from the most relevant affixes in order to enrich the decoder's current state and constrain it to provide better predictions. We evaluated our model to translate English into German, Russian, and Turkish as three MRLs and observed significant improvements.

Alberto Poncelas, Dimitar Shterionov, Andy Way et al.

A prerequisite for training corpus-based machine translation (MT) systems -- either Statistical MT (SMT) or Neural MT (NMT) -- is the availability of high-quality parallel data. This is arguably more important today than ever before, as NMT has been shown in many studies to outperform SMT, but mostly when large parallel corpora are available; in cases where data is limited, SMT can still outperform NMT. Recently researchers have shown that back-translating monolingual data can be used to create synthetic parallel corpora, which in turn can be used in combination with authentic parallel data to train a high-quality NMT system. Given that large collections of new parallel text become available only quite rarely, backtranslation has become the norm when building state-of-the-art NMT systems, especially in resource-poor scenarios. However, we assert that there are many unknown factors regarding the actual effects of back-translated data on the translation capabilities of an NMT model. Accordingly, in this work we investigate how using back-translated data as a training corpus -- both as a separate standalone dataset as well as combined with human-generated parallel data -- affects the performance of an NMT model. We use incrementally larger amounts of back-translated data to train a range of NMT systems for German-to-English, and analyse the resulting translation performance.

Andy Way

Machine Translation (MT) is being deployed for a range of use-cases by millions of people on a daily basis. There should, therefore, be no doubt as to the utility of MT. However, not everyone is convinced that MT can be useful, especially as a productivity enhancer for human translators. In this chapter, I address this issue, describing how MT is currently deployed, how its output is evaluated and how this could be enhanced, especially as MT quality itself improves. Central to these issues is the acceptance that there is no longer a single 'gold standard' measure of quality, such that the situation in which MT is deployed needs to be borne in mind, especially with respect to the expected 'shelf-life' of the translation itself.

Antonio Toral, Andy Way

Given the rise of a new approach to MT, Neural MT (NMT), and its promising performance on different text types, we assess the translation quality it can attain on what is perceived to be the greatest challenge for MT: literary text. Specifically, we target novels, arguably the most popular type of literary text. We build a literary-adapted NMT system for the English-to-Catalan translation direction and evaluate it against a system pertaining to the previous dominant paradigm in MT: statistical phrase-based MT (PBSMT). To this end, for the first time we train MT systems, both NMT and PBSMT, on large amounts of literary text (over 100 million words) and evaluate them on a set of twelve widely known novels spanning from the the 1920s to the present day. According to the BLEU automatic evaluation metric, NMT is significantly better than PBSMT (p < 0.01) on all the novels considered. Overall, NMT results in a 11% relative improvement (3 points absolute) over PBSMT. A complementary human evaluation on three of the books shows that between 17% and 34% of the translations, depending on the book, produced by NMT (versus 8% and 20% with PBSMT) are perceived by native speakers of the target language to be of equivalent quality to translations produced by a professional human translator.

Longyue Wang, Zhaopeng Tu, Andy Way et al.

In translation, considering the document as a whole can help to resolve ambiguities and inconsistencies. In this paper, we propose a cross-sentence context-aware approach and investigate the influence of historical contextual information on the performance of neural machine translation (NMT). First, this history is summarized in a hierarchical way. We then integrate the historical representation into NMT in two strategies: 1) a warm-start of encoder and decoder states, and 2) an auxiliary context source for updating decoder states. Experimental results on a large Chinese-English translation task show that our approach significantly improves upon a strong attention-based NMT system by up to +2.1 BLEU points.

Longyue Wang, Xiaojun Zhang, Zhaopeng Tu et al.

In this paper, a novel approach is proposed to automatically construct parallel discourse corpus for dialogue machine translation. Firstly, the parallel subtitle data and its corresponding monolingual movie script data are crawled and collected from Internet. Then tags such as speaker and discourse boundary from the script data are projected to its subtitle data via an information retrieval approach in order to map monolingual discourse to bilingual texts. We not only evaluate the mapping results, but also integrate speaker information into the translation. Experiments show our proposed method can achieve 81.79% and 98.64% accuracy on speaker and dialogue boundary annotation, and speaker-based language model adaptation can obtain around 0.5 BLEU points improvement in translation qualities. Finally, we publicly release around 100K parallel discourse data with manual speaker and dialogue boundary annotation.

Longyue Wang, Zhaopeng Tu, Xiaojun Zhang et al.

Dropped Pronouns (DP) in which pronouns are frequently dropped in the source language but should be retained in the target language are challenge in machine translation. In response to this problem, we propose a semi-supervised approach to recall possibly missing pronouns in the translation. Firstly, we build training data for DP generation in which the DPs are automatically labelled according to the alignment information from a parallel corpus. Secondly, we build a deep learning-based DP generator for input sentences in decoding when no corresponding references exist. More specifically, the generation is two-phase: (1) DP position detection, which is modeled as a sequential labelling task with recurrent neural networks; and (2) DP prediction, which employs a multilayer perceptron with rich features. Finally, we integrate the above outputs into our translation system to recall missing pronouns by both extracting rules from the DP-labelled training data and translating the DP-generated input sentences. Experimental results show that our approach achieves a significant improvement of 1.58 BLEU points in translation performance with 66% F-score for DP generation accuracy.