Tensor2Tensor, shortly known as T2T, is a library of pre-configured deep learning models and datasets. The Google Brain team has developed it to do deep learning research faster and more accessible. It uses TensorFlow throughout and aims to improve performance and usability strongly. Models can be trained on any of the CPU, single GPU, multiple GPU and TPU either locally or in the cloud. Tensor2Tensor models need minimal or zero configuration or device-specific code. It provides support for well-acclaimed models and datasets across different media platforms such as images, videos, text and audio.

In most machine learning tasks, we evaluate a model $M$ on a given data population $S$ by measuring a population-level metric $F(S;M)$. Examples of such evaluation metric $F$ include precision/recall for (binary) recognition, the F1 score for multi-class classification, and the BLEU metric for language generation. On the other hand, the model $M$ is trained by optimizing a sample-level loss $G(S_t;M)$ at each learning step $t$, where $S_t$ is a subset of $S$ (a.k.a. the mini-batch). Popular choices of $G$ include cross-entropy loss, the Dice loss, and sentence-level BLEU scores. A fundamental assumption behind this paradigm is that the mean value of the sample-level loss $G$, if averaged over all possible samples, should effectively represent the population-level metric $F$ of the task, such as, that $\mathbb{E}[ G(S_t;M) ] \approx F(S;M)$. In this paper, we systematically investigate the above assumption in several NLP tasks. We show, both theoretically and experimentally, that some popular designs of the sample-level loss $G$ may be inconsistent with the true population-level metric $F$ of the task, so that models trained to optimize the former can be substantially sub-optimal to the latter, a phenomenon we call it, Simpson's bias, due to its deep connections with the classic paradox known as Simpson's reversal paradox in statistics and social sciences.

Why do models often attend to salient words, and how does this evolve throughout training? We approximate model training as a two stage process: early on in training when the attention weights are uniform, the model learns to translate individual input word `i` to `o` if they co-occur frequently. Later, the model learns to attend to `i` while the correct output is $o$ because it knows `i` translates to `o`. To formalize, we define a model property, Knowledge to Translate Individual Words (KTIW) (e.g. knowing that `i` translates to `o`), and claim that it drives the learning of the attention. This claim is supported by the fact that before the attention mechanism is learned, KTIW can be learned from word co-occurrence statistics, but not the other way around. Particularly, we can construct a training distribution that makes KTIW hard to learn, the learning of the attention fails, and the model cannot even learn the simple task of copying the input words to the output. Our approximation explains why models sometimes attend to salient words, and inspires a toy example where a multi-head attention model can overcome the above hard training distribution by improving learning dynamics rather than expressiveness.

Obtaining high-quality parallel corpora is of paramount importance for training NMT systems. However, as many language pairs lack adequate gold-standard training data, a popular approach has been to mine so-called "pseudo-parallel" sentences from paired documents in two languages. In this paper, we outline some problems with current methods, propose computationally economical solutions to those problems, and demonstrate success with novel methods on the Tatoeba similarity search benchmark and on a downstream task, namely NMT. We uncover the effect of resource-related factors (i.e. how much monolingual/bilingual data is available for a given language) on the optimal choice of bitext mining approach, and echo problems with the oft-used BUCC dataset that have been observed by others. We make the code and data used for our experiments publicly available.

Recent studies have demonstrated a perceivable improvement on the performance of neural machine translation by applying cross-lingual language model pretraining (Lample and Conneau, 2019), especially the Translation Language Modeling (TLM). To alleviate the need for expensive parallel corpora by TLM, in this work, we incorporate the translation information from dictionaries into the pretraining process and propose a novel Bilingual Dictionary-based Language Model (BDLM). We evaluate our BDLM in Chinese, English, and Romanian. For Chinese-English, we obtained a 55.0 BLEU on WMT-News19 (Tiedemann, 2012) and a 24.3 BLEU on WMT20 news-commentary, outperforming the Vanilla Transformer (Vaswani et al., 2017) by more than 8.4 BLEU and 2.3 BLEU, respectively. According to our results, the BDLM also has advantages on convergence speed and predicting rare words. The increase in BLEU for WMT16 Romanian-English also shows its effectiveness in low-resources language translation.

Translating between languages where certain features are marked morphologically in one but absent or marked contextually in the other is an important test case for machine translation. When translating into English which marks (in)definiteness morphologically, from Yor\`ub\'a which uses bare nouns but marks these features contextually, ambiguities arise. In this work, we perform fine-grained analysis on how an SMT system compares with two NMT systems (BiLSTM and Transformer) when translating bare nouns in Yor\`ub\'a into English. We investigate how the systems what extent they identify BNs, correctly translate them, and compare with human translation patterns. We also analyze the type of errors each model makes and provide a linguistic description of these errors. We glean insights for evaluating model performance in low-resource settings. In translating bare nouns, our results show the transformer model outperforms the SMT and BiLSTM models for 4 categories, the BiLSTM outperforms the SMT model for 3 categories while the SMT outperforms the NMT models for 1 category.

During the initial days, Google Translate was launched with Phrase-Based Machine Translation as the key algorithm. The main improvement in the translation systems was achieved with the introduction of Google Neural Machine Translation or GNMT . With Translatotron, Google demonstrated that a single sequence-to-sequence model can directly translate speech from one language into speech in another language, without the need for intermediate text representation, unlike cascaded systems. Translatotron is claimed to be the first end-to-end model that could directly translate speech from one language into speech in another language and was also able to retain the source speaker's voice in the translated speech. Stay updated on last news about Artificial Intelligence.

Intersectional bias is a bias caused by an overlap of multiple social factors like gender, sexuality, race, disability, religion, etc. A recent study has shown that word embedding models can be laden with biases against intersectional groups like African American females, etc. The first step towards tackling such intersectional biases is to identify them. However, discovering biases against different intersectional groups remains a challenging task. In this work, we present WordBias, an interactive visual tool designed to explore biases against intersectional groups encoded in static word embeddings. Given a pretrained static word embedding, WordBias computes the association of each word along different groups based on race, age, etc. and then visualizes them using a novel interactive interface. Using a case study, we demonstrate how WordBias can help uncover biases against intersectional groups like Black Muslim Males, Poor Females, etc. encoded in word embedding. In addition, we also evaluate our tool using qualitative feedback from expert interviews. The source code for this tool can be publicly accessed for reproducibility at github.com/bhavyaghai/WordBias.

With sequentially stacked self-attention, (optional) encoder-decoder attention, and feed-forward layers, Transformer achieves big success in natural language processing (NLP), and many variants have been proposed. Currently, almost all these models assume that the layer order is fixed and kept the same across data samples. We observe that different data samples actually favor different orders of the layers. Based on this observation, in this work, we break the assumption of the fixed layer order in the Transformer and introduce instance-wise layer reordering into the model structure. Our Instance-wise Ordered Transformer (IOT) can model variant functions by reordered layers, which enables each sample to select the better one to improve the model performance under the constraint of almost the same number of parameters. To achieve this, we introduce a light predictor with negligible parameter and inference cost to decide the most capable and favorable layer order for any input sequence. Experiments on 3 tasks (neural machine translation, abstractive summarization, and code generation) and 9 datasets demonstrate consistent improvements of our method. We further show that our method can also be applied to other architectures beyond Transformer. Our code is released at Github.

Two popular types of machine translation (MT) are phrase-based and neural machine translation systems. Both of these types of systems are composed of multiple complex models or layers. Each of these models and layers learns different linguistic aspects of the source language. However, for some of these models and layers, it is not clear which linguistic phenomena are learned or how this information is learned. For phrase-based MT systems, it is often clear what information is learned by each model, and the question is rather how this information is learned, especially for its phrase reordering model. For neural machine translation systems, the situation is even more complex, since for many cases it is not exactly clear what information is learned and how it is learned. To shed light on what linguistic phenomena are captured by MT systems, we analyze the behavior of important models in both phrase-based and neural MT systems. We consider phrase reordering models from phrase-based MT systems to investigate which words from inside of a phrase have the biggest impact on defining the phrase reordering behavior. Additionally, to contribute to the interpretability of neural MT systems we study the behavior of the attention model, which is a key component in neural MT systems and the closest model in functionality to phrase reordering models in phrase-based systems. The attention model together with the encoder hidden state representations form the main components to encode source side linguistic information in neural MT. To this end, we also analyze the information captured in the encoder hidden state representations of a neural MT system. We investigate the extent to which syntactic and lexical-semantic information from the source side is captured by hidden state representations of different neural MT architectures.