The great majority of languages in the world are considered under-resourced for the successful application of deep learning methods. In this work, we propose a meta-learning approach to document classification in limited-resource setting and demonstrate its effectiveness in two different settings: few-shot, cross-lingual adaptation to previously unseen languages; and multilingual joint training when limited target-language data is available during training. We conduct a systematic comparison of several meta-learning methods, investigate multiple settings in terms of data availability and show that meta-learning thrives in settings with a heterogeneous task distribution. We propose a simple, yet effective adjustment to existing meta-learning methods which allows for better and more stable learning, and set a new state of the art on several languages while performing on-par on others, using only a small amount of labeled data.
Recent advances in cross-lingual word embeddings have primarily relied on mapping-based methods, which project pretrained word embeddings from different languages into a shared space through a linear transformation. However, these approaches assume word embedding spaces are isomorphic between different languages, which has been shown not to hold in practice (S{\o}gaard et al., 2018), and fundamentally limits their performance. This motivates investigating joint learning methods which can overcome this impediment, by simultaneously learning embeddings across languages via a cross-lingual term in the training objective. We propose a bilingual extension of the CBOW method which leverages sentence-aligned corpora to obtain robust cross-lingual word and sentence representations. Our approach significantly improves cross-lingual sentence retrieval performance over all other approaches while maintaining parity with the current state-of-the-art methods on word-translation. It also achieves parity with a deep RNN method on a zero-shot cross-lingual document classification task, requiring far fewer computational resources for training and inference. As an additional advantage, our bilingual method leads to a much more pronounced improvement in the the quality of monolingual word vectors compared to other competing methods.
Word embeddings are high dimensional vector representations of words that capture their semantic similarity in the vector space. There exist several algorithms for learning such embeddings both for a single language as well as for several languages jointly. In this work we propose to evaluate collections of embeddings by adapting downstream natural language tasks to the optimal transport framework. We show how the family of Wasserstein distances can be used to solve cross-lingual document retrieval and the cross-lingual document classification problems. We argue on the advantages of this approach compared to more traditional evaluation methods of embeddings like bilingual lexical induction. Our experimental results suggest that using Wasserstein distances on these problems out-performs several strong baselines and performs on par with state-of-the-art models.
Large deep learning models offer significant accuracy gains, but training billions to trillions of parameters is challenging. Existing solutions such as data and model parallelisms exhibit fundamental limitations to fit these models into limited device memory, while obtaining computation, communication and development efficiency. We develop a novel solution, Zero Redundancy Optimizer (ZeRO), to optimize memory, vastly improving training speed while increasing the model size that can be efficiently trained. ZeRO eliminates memory redundancies in data- and model-parallel training while retaining low communication volume and high computational granularity, allowing us to scale the model size proportional to the number of devices with sustained high efficiency. Our analysis on memory requirements and communication volume demonstrates: ZeRO has the potential to scale beyond 1 Trillion parameters using today's hardware. We implement and evaluate ZeRO: it trains large models of over 100B parameter with super-linear speedup on 400 GPUs, achieving throughput of 15 Petaflops. This represents an 8x increase in model size and 10x increase in achievable performance over state-of-the-art. In terms of usability, ZeRO can train large models of up to 13B parameters (e.g., larger than Megatron GPT 8.3B and T5 11B) without requiring model parallelism which is harder for scientists to apply. Last but not the least, researchers have used the system breakthroughs of ZeRO to create the world's largest language model (Turing-NLG, 17B parameters) with record breaking accuracy.
The scarcity of labeled training data often prohibits the internationalization of NLP models to multiple languages. Recent developments in cross-lingual understanding (XLU) has made progress in this area, trying to bridge the language barrier using language universal representations. However, even if the language problem was resolved, models trained in one language would not transfer to another language perfectly due to the natural domain drift across languages and cultures. We consider the setting of semi-supervised cross-lingual understanding, where labeled data is available in a source language (English), but only unlabeled data is available in the target language. We combine state-of-the-art cross-lingual methods with recently proposed methods for weakly supervised learning such as unsupervised pre-training and unsupervised data augmentation to simultaneously close both the language gap and the domain gap in XLU. We show that addressing the domain gap is crucial. We improve over strong baselines and achieve a new state-of-the-art for cross-lingual document classification.
Pretrained language models are promising particularly for low-resource languages as they only require unlabelled data. However, training existing models requires huge amounts of compute, while pretrained cross-lingual models often underperform on low-resource languages. We propose Multi-lingual language model Fine-Tuning (MultiFiT) to enable practitioners to train and fine-tune language models efficiently in their own language. In addition, we propose a zero-shot method using an existing pretrained cross-lingual model. We evaluate our methods on two widely used cross-lingual classification datasets where they outperform models pretrained on orders of magnitude more data and compute. We release all models and code.
Text classification must sometimes be applied in a low-resource language with no labeled training data. However, training data may be available in a related language. We investigate whether character-level knowledge transfer from a related language helps text classification. We present a cross-lingual document classification framework (CACO) that exploits cross-lingual subword similarity by jointly training a character-based embedder and a word-based classifier. The embedder derives vector representations for input words from their written forms, and the classifier makes predictions based on the word vectors. We use a joint character representation for both the source language and the target language, which allows the embedder to generalize knowledge about source language words to target language words with similar forms. We propose a multi-task objective that can further improve the model if additional cross-lingual or monolingual resources are available. Experiments confirm that character-level knowledge transfer is more data-efficient than word-level transfer between related languages.
We introduce an architecture to learn joint multilingual sentence representations for 93 languages, belonging to more than 30 different families and written in 28 different scripts. Our system uses a single BiLSTM encoder with a shared BPE vocabulary for all languages, which is coupled with an auxiliary decoder and trained on publicly available parallel corpora. This enables us to learn a classifier on top of the resulting embeddings using English annotated data only, and transfer it to any of the 93 languages without any modification. Our experiments in cross-lingual natural language inference (XNLI dataset), cross-lingual document classification (MLDoc dataset) and parallel corpus mining (BUCC dataset) show the effectiveness of our approach. We also introduce a new test set of aligned sentences in 112 languages, and show that our sentence embeddings obtain strong results in multilingual similarity search even for low-resource languages. Our implementation, the pre-trained encoder and the multilingual test set are available at https://github.com/facebookresearch/LASER
This paper investigates a cross-lingual document embedding method that improves the current Neural machine Translation framework based Document Vector (NTDV or simply NV). NV is developed with a self-attention mechanism under the neural machine translation (NMT) framework. In NV, each pair of parallel documents in different languages are projected to the same shared layer in the model. However, the pair of NV embeddings are not guaranteed to be similar. This paper further adds a distance constraint to the training objective function of NV so that the two embeddings of a parallel document are required to be as close as possible. The new method will be called constrained NV (cNV). In a cross-lingual document classification task, the new cNV performs as well as NV and outperforms other published studies that require forward-pass decoding. Compared with the previous NV, cNV does not need a translator during testing, and so the method is lighter and more flexible.
The work investigates deep generative models, which allow us to use training data from one domain to build a model for another domain. We propose the Variational Bi-domain Triplet Autoencoder (VBTA) that learns a joint distribution of objects from different domains. We extend the VBTAs objective function by the relative constraints or triplets that sampled from the shared latent space across domains. In other words, we combine the deep generative models with a metric learning ideas in order to improve the final objective with the triplets information. The performance of the VBTA model is demonstrated on different tasks: image-to-image translation, bi-directional image generation and cross-lingual document classification.
Distributed representations of meaning are a natural way to encode covariance relationships between words and phrases in NLP. By overcoming data sparsity problems, as well as providing information about semantic relatedness which is not available in discrete representations, distributed representations have proven useful in many NLP tasks. Recent work has shown how compositional semantic representations can successfully be applied to a number of monolingual applications such as sentiment analysis. At the same time, there has been some initial success in work on learning shared word-level representations across languages. We combine these two approaches by proposing a method for learning distributed representations in a multilingual setup. Our model learns to assign similar embeddings to aligned sentences and dissimilar ones to sentence which are not aligned while not requiring word alignments. We show that our representations are semantically informative and apply them to a cross-lingual document classification task where we outperform the previous state of the art. Further, by employing parallel corpora of multiple language pairs we find that our model learns representations that capture semantic relationships across languages for which no parallel data was used.
We present a novel technique for learning semantic representations, which extends the distributional hypothesis to multilingual data and joint-space embeddings. Our models leverage parallel data and learn to strongly align the embeddings of semantically equivalent sentences, while maintaining sufficient distance between those of dissimilar sentences. The models do not rely on word alignments or any syntactic information and are successfully applied to a number of diverse languages. We extend our approach to learn semantic representations at the document level, too. We evaluate these models on two cross-lingual document classification tasks, outperforming the prior state of the art. Through qualitative analysis and the study of pivoting effects we demonstrate that our representations are semantically plausible and can capture semantic relationships across languages without parallel data.
In this paper we propose a general framework for learning distributed representations of attributes: characteristics of text whose representations can be jointly learned with word embeddings. Attributes can correspond to document indicators (to learn sentence vectors), language indicators (to learn distributed language representations), meta-data and side information (such as the age, gender and industry of a blogger) or representations of authors. We describe a third-order model where word context and attribute vectors interact multiplicatively to predict the next word in a sequence. This leads to the notion of conditional word similarity: how meanings of words change when conditioned on different attributes. We perform several experimental tasks including sentiment classification, cross-lingual document classification, and blog authorship attribution. We also qualitatively evaluate conditional word neighbours and attribute-conditioned text generation.
We introduce BilBOWA (Bilingual Bag-of-Words without Alignments), a simple and computationally-efficient model for learning bilingual distributed representations of words which can scale to large monolingual datasets and does not require word-aligned parallel training data. Instead it trains directly on monolingual data and extracts a bilingual signal from a smaller set of raw-text sentence-aligned data. This is achieved using a novel sampled bag-of-words cross-lingual objective, which is used to regularize two noise-contrastive language models for efficient cross-lingual feature learning. We show that bilingual embeddings learned using the proposed model outperform state-of-the-art methods on a cross-lingual document classification task as well as a lexical translation task on WMT11 data.
Crosslingual word embeddings represent lexical items from different languages in the same vector space, enabling transfer of NLP tools. However, previous attempts had expensive resource requirements, difficulty incorporating monolingual data or were unable to handle polysemy. We address these drawbacks in our method which takes advantage of a high coverage dictionary in an EM style training algorithm over monolingual corpora in two languages. Our model achieves state-of-the-art performance on bilingual lexicon induction task exceeding models using large bilingual corpora, and competitive results on the monolingual word similarity and cross-lingual document classification task.
In recent years great success has been achieved in sentiment classification for English, thanks in part to the availability of copious annotated resources. Unfortunately, most languages do not enjoy such an abundance of labeled data. To tackle the sentiment classification problem in low-resource languages without adequate annotated data, we propose an Adversarial Deep Averaging Network (ADAN) to transfer the knowledge learned from labeled data on a resource-rich source language to low-resource languages where only unlabeled data exists. ADAN has two discriminative branches: a sentiment classifier and an adversarial language discriminator. Both branches take input from a shared feature extractor to learn hidden representations that are simultaneously indicative for the classification task and invariant across languages. Experiments on Chinese and Arabic sentiment classification demonstrate that ADAN significantly outperforms state-of-the-art systems.
Cross-lingual document classification aims at training a document classifier on resources in one language and transferring it to a different language without any additional resources. Several approaches have been proposed in the literature and the current best practice is to evaluate them on a subset of the Reuters Corpus Volume 2. However, this subset covers only few languages (English, German, French and Spanish) and almost all published works focus on the the transfer between English and German. In addition, we have observed that the class prior distributions differ significantly between the languages. We argue that this complicates the evaluation of the multilinguality. In this paper, we propose a new subset of the Reuters corpus with balanced class priors for eight languages. By adding Italian, Russian, Japanese and Chinese, we cover languages which are very different with respect to syntax, morphology, etc. We provide strong baselines for all language transfer directions using multilingual word and sentence embeddings respectively. Our goal is to offer a freely available framework to evaluate cross-lingual document classification, and we hope to foster by these means, research in this important area.