Persona and Knowledge dual context open-domain chat is a novel dialogue generation task introduced recently. While Persona and Knowledge is each interesting context of open-domain dialogue, the combination of both has not been well studied. We tackle Persona-Knowledge identification and response generation tasks in this paper. We design an informed data augmentation strategy that is compatible with neural Q&A retrieval models. With the augmented data, we perform permutative Persona-Knowledge evaluation and successive Persona search fine-tuning. Furthermore, we perform dialogue generation with various decoding techniques and illustrate crucial elements. We achieve SOTA across official metrics with 93.99% Grounding accuracy average and 23.62 SacreBLEU score.

That's what the doll in Squid Game says. But how would you know! You got subtitles on your plate. Shows like Squid Game and Money Heist topping Netflix charts opened up a whole new genre of drama and entertainment for the audience to explore with different language content. People locked inside the doors during the pandemic brought the world closer together in its unique ways.

The performance of machine translation systems, which previously relied on phrase-based systems, has suddenly improved with the advent of neural network-based models. An open-source framework called Sockeye was released in 2018. This framework provides quick and dependable PyTorch implementation for neural machine translation (NMT) and other related tasks. It supports Amazon Translate and several other NMT applications. In 2020, Sockeye 2, its improved version, was also launched.

Impressive performance of Transformer has been attributed to self-attention, where dependencies between entire input in a sequence are considered at every position. In this work, we reform the neural $n$-gram model, which focuses on only several surrounding representations of each position, with the multi-head mechanism as in Vaswani et al.(2017). Through experiments on sequence-to-sequence tasks, we show that replacing self-attention in Transformer with multi-head neural $n$-gram can achieve comparable or better performance than Transformer. From various analyses on our proposed method, we find that multi-head neural $n$-gram is complementary to self-attention, and their combinations can further improve performance of vanilla Transformer.

We introduce Bi-SimCut: a simple but effective training strategy to boost neural machine translation (NMT) performance. It consists of two procedures: bidirectional pretraining and unidirectional finetuning. Both procedures utilize SimCut, a simple regularization method that forces the consistency between the output distributions of the original and the cutoff sentence pairs. Without leveraging extra dataset via back-translation or integrating large-scale pretrained model, Bi-SimCut achieves strong translation performance across five translation benchmarks (data sizes range from 160K to 20.2M): BLEU scores of 31.16 for en -> de and 38.37 for de -> en on the IWSLT14 dataset, 30.78 for en -> de and 35.15 for de -> en on the WMT14 dataset, and 27.17 for zh -> en on the WMT17 dataset. SimCut is not a new method, but a version of Cutoff (Shen et al., 2020) simplified and adapted for NMT, and it could be considered as a perturbation-based method. Given the universality and simplicity of SimCut and Bi-SimCut, we believe they can serve as strong baselines for future NMT research.

Wouldn't it be great if you could associate IPTC-IM in different languages with a single image? Although the XMP (eXtensibel Metadata Platform) standard image metadata, also known as ISO 16684-1, supports multi-lingual metadata (at least for some fields), currently, none of the commonly used software packages, like Photoshop and Photo Mechanic, support multi-lingual IPTC-IM. Unless your customer uses proprietary data that can read and interpret multi-lingual IPTC-IM, you are bound to manage multiple copies of the same image, only differentiation by the language in which IPTC-IM is written. But a more challenging, or should I say, the time-consuming challenge, is to write IPTC-IM in multiple languages and ensure their consistency, not only when you are not native in each language. At least here on this front, technology can offer a sound solution. It is called AI-based machine translation.

Multi-modal machine translation (MMT) improves translation quality by introducing visual information. However, the existing MMT model ignores the problem that the image will bring information irrelevant to the text, causing much noise to the model and affecting the translation quality. This paper proposes a novel Gumbel-Attention for multi-modal machine translation, which selects the text-related parts of the image features. Specifically, different from the previous attention-based method, we first use a differentiable method to select the image information and automatically remove the useless parts of the image features. Experiments prove that our method retains the image features related to the text, and the remaining parts help the MMT model generates better translations.

Recently, most handwritten mathematical expression recognition (HMER) methods adopt the encoder-decoder networks, which directly predict the markup sequences from formula images with the attention mechanism. However, such methods may fail to accurately read formulas with complicated structure or generate long markup sequences, as the attention results are often inaccurate due to the large variance of writing styles or spatial layouts. To alleviate this problem, we propose an unconventional network for HMER named Counting-Aware Network (CAN), which jointly optimizes two tasks: HMER and symbol counting. Specifically, we design a weakly-supervised counting module that can predict the number of each symbol class without the symbol-level position annotations, and then plug it into a typical attention-based encoder-decoder model for HMER. Experiments on the benchmark datasets for HMER validate that both joint optimization and counting results are beneficial for correcting the prediction errors of encoder-decoder models, and CAN consistently outperforms the state-of-the-art methods. In particular, compared with an encoder-decoder model for HMER, the extra time cost caused by the proposed counting module is marginal. The source code is available at https://github.com/LBH1024/CAN.

Meta AI recently open-sourced its massive translation model, No Language Left Behind (NLLB-200), intending to exclude language barriers across the globe. As we know, that machine translation has become a key area of research nowadays, and it has become a great news for many researchers and organisations who can use it for their respective research and work. So let's take a look at the news and understand a bit about NLLB-200 with the below points: No Language Left Behind (NLLB-200) is a model from the series of massive machine translation models of MetaAI for language translation. A newer member of the series NLLB-200 is capable of translating between 200 languages, representing Meta's capacity of Meta in the direction of AI researchers. These development aims to allow people to access, share and use online content in their native languages and communicate across the world regardless of language preferences.

Recent breakthroughs in Natural Language Processing (NLP) have been driven by language models trained on a massive amount of plain text. While powerful, deriving supervision from textual resources is still an open question. For example, language model pretraining often neglects the rich, freely-available structures in textual data. In this thesis, we describe three lines of work that seek to improve the training and evaluation of neural models using naturally-occurring supervision. We first investigate self-supervised training losses to help enhance the performance of pretrained language models for various NLP tasks. Specifically, we alter the sentence prediction loss to make it better suited to other pretraining losses and more challenging to solve. We design an intermediate finetuning step that uses self-supervised training to promote models' ability in cross-task generalization. Then we describe methods to leverage the structures in Wikipedia and paraphrases. In particular, we propose training losses to exploit hyperlinks, article structures, and article category graphs for entity-, discourse-, entailment-related knowledge. We propose a framework that uses paraphrase pairs to disentangle semantics and syntax in sentence representations. We extend the framework for a novel generation task that controls the syntax of output text with a sentential exemplar. Lastly, we discuss our work on tailoring textual resources for establishing challenging evaluation tasks. We introduce three datasets by defining novel tasks using various fan-contributed websites, including a long-form data-to-text generation dataset, a screenplay summarization dataset, and a long-form story generation dataset. These datasets have unique characteristics offering challenges to future work in their respective task settings.