This paper reports on the shared tasks organized by the 21st IWSLT Conference. The shared tasks address 7 scientific challenges in spoken language translation: simultaneous and offline translation, automatic subtitling and dubbing, speech-to-speech translation, dialect and low-resource speech translation, and Indic languages. The shared tasks attracted 17 teams whose submissions are documented in 27 system papers. The growing interest towards spoken language translation is also witnessed by the constantly increasing number of shared task organizers and contributors to the overview paper, almost evenly distributed across industry and academia.
Thanks to the recent progress in vision-language modeling and the evolving nature of news consumption, the tasks of automatic summarization and headline generation based on multimodal news articles have been gaining popularity. One of the limitations of the current approaches is caused by the commonly used sophisticated modular architectures built upon hierarchical cross-modal encoders and modality-specific decoders, which restrict the model’s applicability to specific data modalities – once trained on, e.g., text+video pairs there is no straightforward way to apply the model to text+image or text-only data. In this work, we propose a unified task formulation that utilizes a simple encoder-decoder model to generate headlines from uni- and multi-modal news articles. This model is trained jointly on data of several modalities and extends the textual decoder to handle the multimodal output.
In recent years, the pattern of news consumption has been changing. The most popular multimedia news formats are now multimodal - the reader is often presented not only with a textual article but also with a short, vivid video. To draw the attention of the reader, such video-based articles are usually presented as a short textual summary paired with an image thumbnail. In this paper, we introduce MLASK (MultimodaL Article Summarization Kit) - a new dataset of video-based news articles paired with a textual summary and a cover picture, all obtained by automatically crawling several news websites. We demonstrate how the proposed dataset can be used to model the task of multimodal summarization by training a Transformer-based neural model. We also examine the effects of pre-training when the usage of generative pre-trained language models helps to improve the model performance, but (additional) pre-training on the simpler task of text summarization yields even better results. Our experiments suggest that the benefits of pre-training and using additional modalities in the input are not orthogonal.
Low-resource Machine Translation (MT) is characterized by the scarce availability of training data and/or standardized evaluation benchmarks. In the context of Dialectal Arabic, recent works introduced several evaluation benchmarks covering both Modern Standard Arabic (MSA) and dialects, mapping, however, mostly to a single Indo-European language - English. In this work, we introduce a multi-lingual corpus consisting of 120,600 multi-parallel sentences in English, French, German, Greek, Spanish, and MSA selected from the OpenSubtitles corpus, which were manually translated into the North Levantine Arabic. By conducting a series of training and fine-tuning experiments, we explore how this novel resource can contribute to the research on Arabic MT.
In this paper, we show that automatically-generated questions and answers can be used to evaluate the quality of Machine Translation (MT) systems. Building on recent work on the evaluation of abstractive text summarization, we propose a new metric for system-level MT evaluation, compare it with other state-of-the-art solutions, and show its robustness by conducting experiments for various MT directions.
In this paper, we describe our submission to the WMT 2021 Metrics Shared Task. We use the automatically-generated questions and answers to evaluate the quality of Machine Translation (MT) systems. Our submission builds upon the recently proposed MTEQA framework. Experiments on WMT20 evaluation datasets show that at the system-level the MTEQA metric achieves performance comparable with other state-of-the-art solutions, while considering only a certain amount of information from the whole translation.
This paper describes the submission to the WMT20 shared news translation task by Samsung R&D Institute Poland. We submitted systems for six language directions: English to Czech, Czech to English, English to Polish, Polish to English, English to Inuktitut and Inuktitut to English. For each, we trained a single-direction model. However, directions including English, Polish and Czech were derived from a common multilingual base, which was later fine-tuned on each particular direction. For all the translation directions, we used a similar training regime, with iterative training corpora improvement through back-translation and model ensembling. For the En → Cs direction, we additionally leveraged document-level information by re-ranking the beam output with a separate model.