Google is helping the Wikimedia Foundation achieve its goal of making Wikipedia articles available in a lot more languages. The Foundation has added Google Translate to its content translation tool, which human editors can use to add content to non-English Wikipedia websites. Those editors can take advantage of the new option -- "one of the most advanced machine translation systems available today," the foundation called it -- to generate an initial translation that they can then review and edit for readability in their language. The Foundation says volunteer Wikipedia editors have been asking for Google Translate integration for a long time now. According to VentureBeat, this move is an expansion of an earlier partnership, wherein Google promised to help Wikipedia make its English posts more accessible in Indonesia.

The study of register in computational language research has historically been divided into register analysis, seeking to determine the registerial character of a text or corpus, and register synthesis, seeking to generate a text in a desired register. This article surveys the different approaches to these disparate tasks. Register synthesis has tended to use more theoretically articulated notions of register and genre than analysis work, which often seeks to categorize on the basis of intuitive and somewhat incoherent notions of prelabeled 'text types'. I argue that an integration of computational register analysis and synthesis will benefit register studies as a whole, by enabling a new large-scale research program in register studies. It will enable comprehensive global mapping of functional language varieties in multiple languages, including the relationships between them. Furthermore, computational methods together with high coverage systematically collected and analyzed data will thus enable rigorous empirical validation and refinement of different theories of register, which will have also implications for our understanding of linguistic variation in general.

The education ministry plans to set up a new subsidy system for prefectures and large cities that offer detailed support to foreign students attending public elementary and junior high schools and their parents through the use of multilingual translation systems. The subsidies will be offered to prefectural governments, ordinance-designated major cities and other core cities that use tablet computers with multilingual speech translation functions when teaching Japanese to students from abroad at school and providing school guidance to their parents. The ministry has set aside ¥20 million for the subsidy system, which is designed to cover one-third of related costs, under the government's fiscal 2019 budget. According to sources, 100 language support programs are likely to become eligible for the financial aid. The launch of the new subsidy system comes in line with the government's policy of allowing more foreign workers to enter the country.

The education ministry plans to set up a new subsidy system for prefectures and large cities that offer detailed support to foreign students attending public elementary and junior high schools and their parents by using multilingual translation systems. The subsidies will be offered to prefectural governments, ordinance-designated major cities and other core cities that use tablet computers with multilingual speech translation functions in teaching Japanese to students from abroad at school and providing school guidance to their parents. The ministry has set aside ¥20 million for the subsidy system, which is designed to cover one-third of related costs, under the government's fiscal 2019 budget, with 100 language support programs likely to become eligible for the financial aid, informed sources said. The launch of the new subsidy system comes in line with the government's policy of allowing more foreign workers to come here. The number of foreign students in Japan needing Japanese language education totaled 43,947 in fiscal 2016, up 70 percent from 26,281 in fiscal 2006.

We investigate the feasibility of learning from both fully-labeled supervised data and contextual bandit data. We specifically consider settings in which the underlying learning signal may be different between these two data sources. Theoretically, we state and prove no-regret algorithms for learning that is robust to divergences between the two sources. Empirically, we evaluate some of these algorithms on a large selection of datasets, showing that our approaches are feasible, and helpful in practice.

Unsupervised image-to-image translation has gained considerable attention due to the recent impressive progress based on generative adversarial networks (GANs). However, previous methods often fail in challenging cases, in particular, when an image has multiple target instances and a translation task involves significant changes in shape, e.g., translating pants to skirts in fashion images. To tackle the issues, we propose a novel method, coined instance-aware GAN (InstaGAN), that incorporates the instance information (e.g., object segmentation masks) and improves multi-instance transfiguration. The proposed method translates both an image and the corresponding set of instance attributes while maintaining the permutation invariance property of the instances. To this end, we introduce a context preserving loss that encourages the network to learn the identity function outside of target instances. We also propose a sequential mini-batch inference/training technique that handles multiple instances with a limited GPU memory and enhances the network to generalize better for multiple instances. Our comparative evaluation demonstrates the effectiveness of the proposed method on different image datasets, in particular, in the aforementioned challenging cases. Code and results are available in https://github.com/sangwoomo/instagan

The future needs to be more human (and less machine). For decades now, we have created computer programming languages and forced entire generations across the globe into becoming engineers and learning how to code. We have succeeded beyond our wildest dreams (we've created machines that can now learn on their own), and we have failed beyond our worst nightmares (we've created "black box" artificial intelligence (AI) which we don't -- and can't -- understand). It's time for us to rethink the future we're so effectively creating. I'm concerned by this trend, where we focus our mental energy into machines, rather than having them understand us more.

In the past few years, artificial intelligence has advanced so quickly that it now seems hardly a month goes by without a newsworthy AI breakthrough. In areas as wide-ranging as speech translation, medical diagnosis, and gameplay, we have seen computers outperform humans in startling ways. This has sparked a discussion about how AI will impact employment. Some fear that as AI improves, it will supplant workers, creating an ever-growing pool of unemployable humans who cannot compete economically with machines. This concern, while understandable, is unfounded.

This is a machine translation reading list maintained by the Tsinghua Natural Language Processing Group. The past three decades have witnessed the rapid development of machine translation, especially for data-driven approaches such as statistical machine translation (SMT) and neural machine translation (NMT). Due to the dominance of NMT at the present time, priority is given to collecting important, up-to-date NMT papers. The list is still incomplete and the categorization might be inappropriate. We will keep adding papers and improving the list.

Binary classifiers are employed as discriminators in GAN-based unsupervised style transfer models to ensure that transferred sentences are similar to sentences in the target domain. One difficulty with the binary discriminator is that error signal is sometimes insufficient to train the model to produce rich-structured language. In this paper, we propose a technique of using a target domain language model as the discriminator to provide richer, token-level feedback during the learning process. Because our language model scores sentences directly using a product of locally normalized probabilities, it offers more stable and more useful training signal to the generator. We train the generator to minimize the negative log likelihood (NLL) of generated sentences evaluated by a language model. By using continuous approximation of the discrete samples, our model can be trained using back-propagation in an end-to-end way. Moreover, we find empirically with a language model as a structured discriminator, it is possible to eliminate the adversarial training steps using negative samples, thus making training more stable. We compare our model with previous work using convolutional neural networks (CNNs) as discriminators and show our model outperforms them significantly in three tasks including word substitution decipherment, sentiment modification and related language translation.