We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. 
BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables networks to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range of prior research has investigated the spatial component of this relationship, seeking to strengthen the representational power of a CNN by enhancing the quality of spatial encodings throughout its feature hierarchy. In this work, we focus instead on the channel relationship and propose a novel architectural unit, which we term the “Squeeze-and-Excitation” (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We show that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. We further demonstrate that SE blocks bring significant improvements in performance for existing state-of-the-art CNNs at slight additional computational cost. Squeeze-and-Excitation Networks formed the foundation of our ILSVRC 2017 classification submission which won first place and reduced the top-5 error to 2.251 percent, surpassing the winning entry of 2016 by a relative improvement of ${\sim }$ ∼ 25 percent. Models and code are available at https://github.com/hujie-frank/SENet .

/pdf/squeeze-and-excitation-networks-50icgwqgeu.pdf

Squeeze-and-Excitation Networks

Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code.

/pdf/roberta-a-robustly-optimized-bert-pretraining-approach-2rj0bdyhim.pdf

RoBERTa: A Robustly Optimized BERT Pretraining Approach

While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

In statistical language modeling, one technique to reduce the problematic eects of data sparsity is to partition the vocabulary into equivalence classes. In this paper we investigate the eects of applying such a technique to higherorder n-gram models trained on large corpora. We introduce a modification of the exchange clustering algorithm with improved eciency for certain partially class-based models and a distributed version of this algorithm to eciently obtain automatic word classifications for large vocabularies (>1 million words) using such large training corpora (>30 billion tokens). The resulting clusterings are then used in training partially class-based language models. We show that combining them with wordbased n-gram models in the log-linear model of a state-of-the-art statistical machine translation system leads to improvements in translation quality as indicated by the BLEU score.

Distributed Word Clustering for Large Scale Class-Based Language Modeling in Machine Translation

We present KERMIT, a simple insertion-based approach to generative modeling for sequences and sequence pairs. KERMIT models the joint distribution and its decompositions (i.e., marginals and conditionals) using a single neural network and, unlike much prior work, does not rely on a prespecified factorization of the data distribution. During training, one can feed KERMIT paired data $(x, y)$ to learn the joint distribution $p(x, y)$, and optionally mix in unpaired data $x$ or $y$ to refine the marginals $p(x)$ or $p(y)$. During inference, we have access to the conditionals $p(x \mid y)$ and $p(y \mid x)$ in both directions. We can also sample from the joint distribution or the marginals. The model supports both serial fully autoregressive decoding and parallel partially autoregressive decoding, with the latter exhibiting an empirically logarithmic runtime. We demonstrate through experiments in machine translation, representation learning, and zero-shot cloze question answering that our unified approach is capable of matching or exceeding the performance of dedicated state-of-the-art systems across a wide range of tasks without the need for problem-specific architectural adaptation.

KERMIT: Generative Insertion-Based Modeling for Sequences

/pdf/fast-decoding-in-sequence-models-using-discrete-latent-4pvc7npi15.pdf

Fast decoding in sequence models using discrete latent variables

Music relies heavily on repetition to build structure and meaning. Self-reference occurs on multiple timescales, from motifs to phrases to reusing of entire sections of music, such as in pieces with ABA structure. The Transformer (Vaswani et al., 2017), a sequence model based on self-attention, has achieved compelling results in many generation tasks that require maintaining long-range coherence. This suggests that self-attention might also be well-suited to modeling music. In musical composition and performance, however, relative timing is critically important. Existing approaches for representing relative positional information in the Transformer modulate attention based on pairwise distance (Shaw et al., 2018). This is impractical for long sequences such as musical compositions since their memory complexity for intermediate relative information is quadratic in the sequence length. We propose an algorithm that reduces their intermediate memory requirement to linear in the sequence length. This enables us to demonstrate that a Transformer with our modified relative attention mechanism can generate minute-long compositions (thousands of steps, four times the length modeled in Oore et al., 2018) with compelling structure, generate continuations that coherently elaborate on a given motif, and in a seq2seq setup generate accompaniments conditioned on melodies. We evaluate the Transformer with our relative attention mechanism on two datasets, JSB Chorales and Piano-e-Competition, and obtain state-of-the-art results on the latter.

Music Transformer

When translating among languages that differ substantially in word order, machine translation (MT) systems benefit from syntactic pre-ordering---an approach that uses features from a syntactic parse to permute source words into a target-language-like order. This paper presents a method for inducing parse trees automatically from a parallel corpus, instead of using a supervised parser trained on a tree-bank. These induced parses are used to pre-order source sentences. We demonstrate that our induced parser is effective: it not only improves a state-of-the-art phrase-based system with integrated reordering, but also approaches the performance of a recent pre-ordering method based on a supervised parser. These results show that the syntactic structure which is relevant to MT pre-ordering can be learned automatically from parallel text, thus establishing a new application for unsupervised grammar induction.

/pdf/inducing-sentence-structure-from-parallel-corpora-for-3bszu0kvmu.pdf

Jakob Uszkoreit

Papers

Distributed Word Clustering for Large Scale Class-Based Language Modeling in Machine Translation

KERMIT: Generative Insertion-Based Modeling for Sequences

Fast decoding in sequence models using discrete latent variables

Music Transformer

Inducing Sentence Structure from Parallel Corpora for Reordering