Pre-trained language models have been shown to improve performance in many natural language tasks substantially. Although the early focus of such models was single language pre-training, recent advances have resulted in cross-lingual and visual pre-training methods. In this paper, we combine these two approaches to learn visually-grounded cross-lingual representations. Specifically, we extend the translation language modelling (Lample and Conneau, 2019) with masked region classification and perform pre-training with three-way parallel vision & language corpora. We show that when fine-tuned for multimodal machine translation, these models obtain state-of-the-art performance. We also provide qualitative insights into the usefulness of the learned grounded representations.
Visual context provides grounding information for multimodal machine translation (MMT). However, previous MMT models and probing studies on visual features suggest that visual information is less explored in MMT as it is often redundant to textual information. In this paper, we propose an object-level visual context modeling framework (OVC) to efficiently capture and explore visual information for multimodal machine translation. With detected objects, the proposed OVC encourages MMT to ground translation on desirable visual objects by masking irrelevant objects in the visual modality. We equip the proposed with an additional object-masking loss to achieve this goal. The object-masking loss is estimated according to the similarity between masked objects and the source texts so as to encourage masking source-irrelevant objects. Additionally, in order to generate vision-consistent target words, we further propose a vision-weighted translation loss for OVC. Experiments on MMT datasets demonstrate that the proposed OVC model outperforms state-of-the-art MMT models and analyses show that masking irrelevant objects helps grounding in MMT.
Automatic generation of video descriptions in natural language, also called video captioning, aims to understand the visual content of the video and produce a natural language sentence depicting the objects and actions in the scene. This challenging integrated vision and language problem, however, has been predominantly addressed for English. The lack of data and the linguistic properties of other languages limit the success of existing approaches for such languages. In this paper we target Turkish, a morphologically rich and agglutinative language that has very different properties compared to English. To do so, we create the first large scale video captioning dataset for this language by carefully translating the English descriptions of the videos in the MSVD (Microsoft Research Video Description Corpus) dataset into Turkish. In addition to enabling research in video captioning in Turkish, the parallel English-Turkish descriptions also enables the study of the role of video context in (multimodal) machine translation. In our experiments, we build models for both video captioning and multimodal machine translation and investigate the effect of different word segmentation approaches and different neural architectures to better address the properties of Turkish. We hope that the MSVD-Turkish dataset and the results reported in this work will lead to better video captioning and multimodal machine translation models for Turkish and other morphology rich and agglutinative languages.
There are common semantics shared across text and images. Given a sentence in a source language, whether depicting the visual scene helps translation into a target language? Existing multimodal neural machine translation methods (MNMT) require triplets of bilingual sentence - image for training and tuples of source sentence - image for inference. In this paper, we propose ImagiT, a novel machine translation method via visual imagination. ImagiT first learns to generate visual representation from the source sentence, and then utilizes both source sentence and the "imagined representation" to produce a target translation. Unlike previous methods, it only needs the source sentence at the inference time. Experiments demonstrate that ImagiT benefits from visual imagination and significantly outperforms the text-only neural machine translation baselines. Further analysis reveals that the imagination process in ImagiT helps fill in missing information when performing the degradation strategy.
Multimodal machine translation (MMT), which mainly focuses on enhancing text-only translation with visual features, has attracted considerable attention from both computer vision and natural language processing communities. Most current MMT models resort to attention mechanism, global context modeling or multimodal joint representation learning to utilize visual features. However, the attention mechanism lacks sufficient semantic interactions between modalities while the other two provide fixed visual context, which is unsuitable for modeling the observed variability when generating translation. To address the above issues, in this paper, we propose a novel Dynamic Context-guided Capsule Network (DCCN) for MMT. Specifically, at each timestep of decoding, we first employ the conventional source-target attention to produce a timestep-specific source-side context vector. Next, DCCN takes this vector as input and uses it to guide the iterative extraction of related visual features via a context-guided dynamic routing mechanism. Particularly, we represent the input image with global and regional visual features, we introduce two parallel DCCNs to model multimodal context vectors with visual features at different granularities. Finally, we obtain two multimodal context vectors, which are fused and incorporated into the decoder for the prediction of the target word. Experimental results on the Multi30K dataset of English-to-German and English-to-French translation demonstrate the superiority of DCCN. Our code is available on https://github.com/DeepLearnXMU/MM-DCCN.
The generalization capability of deep neural networks has been substantially improved by applying a wide spectrum of regularization methods, e.g., restricting function space, injecting randomness during training, augmenting data, etc. In this work, we propose a simple yet effective regularization method named progressive self-knowledge distillation (PS-KD), which progressively distills a model's own knowledge to soften hard targets (i.e., one-hot vectors) during training. Hence, it can be interpreted within a framework of knowledge distillation as a student becomes a teacher itself. Specifically, targets are adjusted adaptively by combining the ground-truth and past predictions from the model itself. We show that PS-KD provides an effect of hard example mining by rescaling gradients according to difficulty in classifying examples. The proposed method is applicable to any supervised learning tasks with hard targets and can be easily combined with existing regularization methods to further enhance the generalization performance. Furthermore, it is confirmed that PS-KD achieves not only better accuracy, but also provides high quality of confidence estimates in terms of calibration as well as ordinal ranking. Extensive experimental results on three different tasks, image classification, object detection, and machine translation, demonstrate that our method consistently improves the performance of the state-of-the-art baselines. The code is available at https://github.com/lgcnsai/PS-KD-Pytorch.
We present M3P, a Multitask Multilingual Multimodal Pre-trained model that combines multilingual pre-training and multimodal pre-training into a unified framework via multitask pre-training. Our goal is to learn universal representations that can map objects occurred in different modalities or texts expressed in different languages into a common semantic space. In addition, to explicitly encourage fine-grained alignment between images and non-English languages, we also propose Multimodal Code-switched Training (MCT) to combine monolingual pre-training and multimodal pre-training via a code-switch strategy. Experiments are performed on the multilingual image retrieval task across two benchmark datasets, including MSCOCO and Multi30K. M3P can achieve comparable results for English and new state-of-the-art results for non-English languages.
We demonstrate how we can practically incorporate multi-step future information into a decoder of maximum likelihood sequence models. We propose a "k-step look-ahead" module to consider the likelihood information of a rollout up to k steps. Unlike other approaches that need to train another value network to evaluate the rollouts, we can directly apply this look-ahead module to improve the decoding of any sequence model trained in a maximum likelihood framework. We evaluate our look-ahead module on three datasets of varying difficulties: IM2LATEX-100k OCR image to LaTeX, WMT16 multimodal machine translation, and WMT14 machine translation. Our look-ahead module improves the performance of the simpler datasets such as IM2LATEX-100k and WMT16 multimodal machine translation. However, the improvement of the more difficult dataset (e.g., containing longer sequences), WMT14 machine translation, becomes marginal. Our further investigation using the k-step look-ahead suggests that the more difficult tasks suffer from the overestimated EOS (end-of-sentence) probability. We argue that the overestimated EOS probability also causes the decreased performance of beam search when increasing its beam width. We tackle the EOS problem by integrating an auxiliary EOS loss into the training to estimate if the model should emit EOS or other words. Our experiments show that improving EOS estimation not only increases the performance of our proposed look-ahead module but also the robustness of the beam search.
Multimodal machine translation involves drawing information from more than one modality, based on the assumption that the additional modalities will contain useful alternative views of the input data. The most prominent tasks in this area are spoken language translation, image-guided translation, and video-guided translation, which exploit audio and visual modalities, respectively. These tasks are distinguished from their monolingual counterparts of speech recognition, image captioning, and video captioning by the requirement of models to generate outputs in a different language. This survey reviews the major data resources for these tasks, the evaluation campaigns concentrated around them, the state of the art in end-to-end and pipeline approaches, and also the challenges in performance evaluation. The paper concludes with a discussion of directions for future research in these areas: the need for more expansive and challenging datasets, for targeted evaluations of model performance, and for multimodality in both the input and output space.
Effective fusion of data from multiple modalities, such as video, speech, and text, is challenging due to the heterogeneous nature of multimodal data. In this paper, we propose adaptive fusion techniques that aim to model context from different modalities effectively. Instead of defining a deterministic fusion operation, such as concatenation, for the network, we let the network decide "how" to combine a given set of multimodal features more effectively. We propose two networks: 1) Auto-Fusion, which learns to compress information from different modalities while preserving the context, and 2) GAN-Fusion, which regularizes the learned latent space given context from complementing modalities. A quantitative evaluation on the tasks of multimodal machine translation and emotion recognition suggests that our lightweight, adaptive networks can better model context from other modalities than existing methods, many of which employ massive transformer-based networks.
This paper describes the cascaded multimodal speech translation systems developed by Imperial College London for the IWSLT 2019 evaluation campaign. The architecture consists of an automatic speech recognition (ASR) system followed by a Transformer-based multimodal machine translation (MMT) system. While the ASR component is identical across the experiments, the MMT model varies in terms of the way of integrating the visual context (simple conditioning vs. attention), the type of visual features exploited (pooled, convolutional, action categories) and the underlying architecture. For the latter, we explore both the canonical transformer and its deliberation version with additive and cascade variants which differ in how they integrate the textual attention. Upon conducting extensive experiments, we found that (i) the explored visual integration schemes often harm the translation performance for the transformer and additive deliberation, but considerably improve the cascade deliberation; (ii) the transformer and cascade deliberation integrate the visual modality better than the additive deliberation, as shown by the incongruence analysis.
Leveraging the visual modality effectively for Neural Machine Translation (NMT) remains an open problem in computational linguistics. Recently, Caglayan et al. posit that the observed gains are limited mainly due to the very simple, short, repetitive sentences of the Multi30k dataset (the only multimodal MT dataset available at the time), which renders the source text sufficient for context. In this work, we further investigate this hypothesis on a new large scale multimodal Machine Translation (MMT) dataset, How2, which has 1.57 times longer mean sentence length than Multi30k and no repetition. We propose and evaluate three novel fusion techniques, each of which is designed to ensure the utilization of visual context at different stages of the Sequence-to-Sequence transduction pipeline, even under full linguistic context. However, we still obtain only marginal gains under full linguistic context and posit that visual embeddings extracted from deep vision models (ResNet for Multi30k, ResNext for How2) do not lend themselves to increasing the discriminativeness between the vocabulary elements at token level prediction in NMT. We demonstrate this qualitatively by analyzing attention distribution and quantitatively through Principal Component Analysis, arriving at the conclusion that it is the quality of the visual embeddings rather than the length of sentences, which need to be improved in existing MMT datasets.
Recent literature shows that large-scale language modeling provides excellent reusable sentence representations with both recurrent and self-attentive architectures. However, there has been less clarity on the commonalities and differences in the representational properties induced by the two architectures. It also has been shown that visual information serves as one of the means for grounding sentence representations. In this paper, we present a meta-study assessing the representational quality of models where the training signal is obtained from different modalities, in particular, language modeling, image features prediction, and both textual and multimodal machine translation. We evaluate textual and visual features of sentence representations obtained using predominant approaches on image retrieval and semantic textual similarity. Our experiments reveal that on moderate-sized datasets, a sentence counterpart in a target language or visual modality provides much stronger training signal for sentence representation than language modeling. Importantly, we observe that while the Transformer models achieve superior machine translation quality, representations from the recurrent neural network based models perform significantly better over tasks focused on semantic relevance.
Visual Genome is a dataset connecting structured image information with English language. We present ``Hindi Visual Genome'', a multimodal dataset consisting of text and images suitable for English-Hindi multimodal machine translation task and multimodal research. We have selected short English segments (captions) from Visual Genome along with associated images and automatically translated them to Hindi with manual post-editing which took the associated images into account. We prepared a set of 31525 segments, accompanied by a challenge test set of 1400 segments. This challenge test set was created by searching for (particularly) ambiguous English words based on the embedding similarity and manually selecting those where the image helps to resolve the ambiguity. Our dataset is the first for multimodal English-Hindi machine translation, freely available for non-commercial research purposes. Our Hindi version of Visual Genome also allows to create Hindi image labelers or other practical tools. Hindi Visual Genome also serves in Workshop on Asian Translation (WAT) 2019 Multi-Modal Translation Task.
Previous work on multimodal machine translation has shown that visual information is only needed in very specific cases, for example in the presence of ambiguous words where the textual context is not sufficient. As a consequence, models tend to learn to ignore this information. We propose a translate-and-refine approach to this problem where images are only used by a second stage decoder. This approach is trained jointly to generate a good first draft translation and to improve over this draft by (i) making better use of the target language textual context (both left and right-side contexts) and (ii) making use of visual context. This approach leads to the state of the art results. Additionally, we show that it has the ability to recover from erroneous or missing words in the source language.
In recent years, pretrained word embeddings have proved useful for multimodal neural machine translation (NMT) models to address the shortage of available datasets. However, the integration of pretrained word embeddings has not yet been explored extensively. Further, pretrained word embeddings in high dimensional spaces have been reported to suffer from the hubness problem. Although some debiasing techniques have been proposed to address this problem for other natural language processing tasks, they have seldom been studied for multimodal NMT models. In this study, we examine various kinds of word embeddings and introduce two debiasing techniques for three multimodal NMT models and two language pairs -- English-German translation and English-French translation. With our optimal settings, the overall performance of multimodal models was improved by up to +1.93 BLEU and +2.02 METEOR for English-German translation and +1.73 BLEU and +0.95 METEOR for English-French translation.
Multimodal machine translation is an attractive application of neural machine translation (NMT). It helps computers to deeply understand visual objects and their relations with natural languages. However, multimodal NMT systems suffer from a shortage of available training data, resulting in poor performance for translating rare words. In NMT, pretrained word embeddings have been shown to improve NMT of low-resource domains, and a search-based approach is proposed to address the rare word problem. In this study, we effectively combine these two approaches in the context of multimodal NMT and explore how we can take full advantage of pretrained word embeddings to better translate rare words. We report overall performance improvements of 1.24 METEOR and 2.49 BLEU and achieve an improvement of 7.67 F-score for rare word translation.
Current work on multimodal machine translation (MMT) has suggested that the visual modality is either unnecessary or only marginally beneficial. We posit that this is a consequence of the very simple, short and repetitive sentences used in the only available dataset for the task (Multi30K), rendering the source text sufficient as context. In the general case, however, we believe that it is possible to combine visual and textual information in order to ground translations. In this paper we probe the contribution of the visual modality to state-of-the-art MMT models by conducting a systematic analysis where we partially deprive the models from source-side textual context. Our results show that under limited textual context, models are capable of leveraging the visual input to generate better translations. This contradicts the current belief that MMT models disregard the visual modality because of either the quality of the image features or the way they are integrated into the model.
In this work, we propose to model the interaction between visual and textual features for multi-modal neural machine translation (MMT) through a latent variable model. This latent variable can be seen as a multi-modal stochastic embedding of an image and its description in a foreign language. It is used in a target-language decoder and also to predict image features. Importantly, our model formulation utilises visual and textual inputs during training but does not require that images be available at test time. We show that our latent variable MMT formulation improves considerably over strong baselines, including a multi-task learning approach (Elliott and K\'ad\'ar, 2017) and a conditional variational auto-encoder approach (Toyama et al., 2016). Finally, we show improvements due to (i) predicting image features in addition to only conditioning on them, (ii) imposing a constraint on the minimum amount of information encoded in the latent variable, and (iii) by training on additional target-language image descriptions (i.e. synthetic data).
This paper describes the UMONS solution for the Multimodal Machine Translation Task presented at the third conference on machine translation (WMT18). We explore a novel architecture, called deepGRU, based on recent findings in the related task of Neural Image Captioning (NIC). The models presented in the following sections lead to the best METEOR translation score for both constrained (English, image) -> German and (English, image) -> French sub-tasks.