We introduce a variety of models, trained on a supervised image captioning corpus to predict the image features for a given caption, to perform sentence representation grounding. We train a grounded sentence encoder that achieves good performance on COCO caption and image retrieval and subsequently show that this encoder can successfully be transferred to various NLP tasks, with improved performance over text-only models. Lastly, we analyze the contribution of grounding, and show that word embeddings learned by this system outperform non-grounded ones.
Zero-shot learning (ZSL) aims to discriminate images from unseen classes by exploiting relations to seen classes via their semantic descriptions. Some recent papers have shown the importance of localized features together with fine-tuning the feature extractor to obtain discriminative and transferable features. However, these methods require complex attention or part detection modules to perform explicit localization in the visual space. In contrast, in this paper we propose localizing representations in the semantic/attribute space, with a simple but effective pipeline where localization is implicit. Focusing on attribute representations, we show that our method obtains state-of-the-art performance on CUB and SUN datasets, and also achieves competitive results on AWA2 dataset, outperforming generally more complex methods with explicit localization in the visual space. Our method can be implemented easily, which can be used as a new baseline for zero shot learning.
This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels.
This paper shows that pretraining multilingual language models at scale leads to significant performance gains for a wide range of cross-lingual transfer tasks. We train a Transformer-based masked language model on one hundred languages, using more than two terabytes of filtered CommonCrawl data. Our model, dubbed XLM-R, significantly outperforms multilingual BERT (mBERT) on a variety of cross-lingual benchmarks, including +13.8% average accuracy on XNLI, +12.3% average F1 score on MLQA, and +2.1% average F1 score on NER. XLM-R performs particularly well on low-resource languages, improving 11.8% in XNLI accuracy for Swahili and 9.2% for Urdu over the previous XLM model. We also present a detailed empirical evaluation of the key factors that are required to achieve these gains, including the trade-offs between (1) positive transfer and capacity dilution and (2) the performance of high and low resource languages at scale. Finally, we show, for the first time, the possibility of multilingual modeling without sacrificing per-language performance; XLM-Ris very competitive with strong monolingual models on the GLUE and XNLI benchmarks. We will make XLM-R code, data, and models publicly available.
For many computer vision applications such as image captioning, visual question answering, and person search, learning discriminative feature representations at both image and text level is an essential yet challenging problem. Its challenges originate from the large word variance in the text domain as well as the difficulty of accurately measuring the distance between the features of the two modalities. Most prior work focuses on the latter challenge, by introducing loss functions that help the network learn better feature representations but fail to account for the complexity of the textual input. With that in mind, we introduce TIMAM: a Text-Image Modality Adversarial Matching approach that learns modality-invariant feature representations using adversarial and cross-modal matching objectives. In addition, we demonstrate that BERT, a publicly-available language model that extracts word embeddings, can successfully be applied in the text-to-image matching domain. The proposed approach achieves state-of-the-art cross-modal matching performance on four widely-used publicly-available datasets resulting in absolute improvements ranging from 2% to 5% in terms of rank-1 accuracy.
Pre-trained language model representations have been successful in a wide range of language understanding tasks. In this paper, we examine different strategies to integrate pre-trained representations into sequence to sequence models and apply it to neural machine translation and abstractive summarization. We find that pre-trained representations are most effective when added to the encoder network which slows inference by only 14%. Our experiments in machine translation show gains of up to 5.3 BLEU in a simulated resource-poor setup. While returns diminish with more labeled data, we still observe improvements when millions of sentence-pairs are available. Finally, on abstractive summarization we achieve a new state of the art on the full text version of CNN/DailyMail.
In this paper, we propose a method for obtaining sentence-level embeddings. While the problem of securing word-level embeddings is very well studied, we propose a novel method for obtaining sentence-level embeddings. This is obtained by a simple method in the context of solving the paraphrase generation task. If we use a sequential encoder-decoder model for generating paraphrase, we would like the generated paraphrase to be semantically close to the original sentence. One way to ensure this is by adding constraints for true paraphrase embeddings to be close and unrelated paraphrase candidate sentence embeddings to be far. This is ensured by using a sequential pair-wise discriminator that shares weights with the encoder that is trained with a suitable loss function. Our loss function penalizes paraphrase sentence embedding distances from being too large. This loss is used in combination with a sequential encoder-decoder network. We also validated our method by evaluating the obtained embeddings for a sentiment analysis task. The proposed method results in semantic embeddings and outperforms the state-of-the-art on the paraphrase generation and sentiment analysis task on standard datasets. These results are also shown to be statistically significant.
Most existing sentiment analysis approaches heavily rely on a large amount of labeled data that usually involve time-consuming and error-prone manual annotations. The distribution of this labeled data is significantly imbalanced among languages, e.g., more English texts are labeled than texts in other languages, which presents a major challenge to cross-lingual sentiment analysis. There have been several cross-lingual representation learning techniques that transfer the knowledge learned from a language with abundant labeled examples to another language with much fewer labels. Their performance, however, is usually limited due to the imperfect quality of machine translation and the scarce signal that bridges two languages. In this paper, we employ emojis, a ubiquitous and emotional language, as a new bridge for sentiment analysis across languages. Specifically, we propose a semi-supervised representation learning approach through the task of emoji prediction to learn cross-lingual representations of text that can capture both semantic and sentiment information. The learned representations are then utilized to facilitate cross-lingual sentiment classification. We demonstrate the effectiveness and efficiency of our approach on a representative Amazon review data set that covers three languages and three domains.
Fine-grained image classification is to recognize hundreds of subcategories belonging to the same basic-level category, which is a highly challenging task due to the quite subtle visual distinctions among similar subcategories. Most existing methods generally learn part detectors to discover discriminative regions for better performance. However, not all localized parts are beneficial and indispensable for classification, and the setting for number of part detectors relies heavily on prior knowledge as well as experimental results. As is known to all, when we describe the object of an image into text via natural language, we only focus on the pivotal characteristics, and rarely pay attention to common characteristics as well as the background areas. This is an involuntary transfer from human visual attention to textual attention, which leads to the fact that textual attention tells us how many and which parts are discriminative and significant. So textual attention of natural language descriptions could help us to discover visual attention in image. Inspired by this, we propose a visual-textual attention driven fine-grained representation learning (VTA) approach, and its main contributions are: (1) Fine-grained visual-textual pattern mining devotes to discovering discriminative visual-textual pairwise information for boosting classification through jointly modeling vision and text with generative adversarial networks (GANs), which automatically and adaptively discovers discriminative parts. (2) Visual-textual representation learning jointly combine visual and textual information, which preserves the intra-modality and inter-modality information to generate complementary fine-grained representation, and further improve classification performance. Experiments on the two widely-used datasets demonstrate the effectiveness of our VTA approach, which achieves the best classification accuracy.
A visual-relational knowledge graph (KG) is a multi-relational graph whose entities are associated with images. We introduce ImageGraph, a KG with 1,330 relation types, 14,870 entities, and 829,931 images. Visual-relational KGs lead to novel probabilistic query types where images are treated as first-class citizens. Both the prediction of relations between unseen images and multi-relational image retrieval can be formulated as query types in a visual-relational KG. We approach the problem of answering such queries with a novel combination of deep convolutional networks and models for learning knowledge graph embeddings. The resulting models can answer queries such as "How are these two unseen images related to each other?" We also explore a zero-shot learning scenario where an image of an entirely new entity is linked with multiple relations to entities of an existing KG. The multi-relational grounding of unseen entity images into a knowledge graph serves as the description of such an entity. We conduct experiments to demonstrate that the proposed deep architectures in combination with KG embedding objectives can answer the visual-relational queries efficiently and accurately.
Sentence representation models trained only on language could potentially suffer from the grounding problem. Recent work has shown promising results in improving the qualities of sentence representations by jointly training them with associated image features. However, the grounding capability is limited due to distant connection between input sentences and image features by the design of the architecture. In order to further close the gap, we propose applying self-attention mechanism to the sentence encoder to deepen the grounding effect. Our results on transfer tasks show that self-attentive encoders are better for visual grounding, as they exploit specific words with strong visual associations.