Vast requirement of computation power of Deep Neural Networks is a major hurdle to their real world applications. Many recent Application Specific Integrated Circuit (ASIC) chips feature dedicated hardware support for Neural Network Acceleration. However, as ASICs take multiple years to develop, they are inevitably out-paced by the latest development in Neural Architecture Research. For example, Transformer Networks do not have native support on many popular chips, and hence are difficult to deploy. In this paper, we propose Arch-Net, a family of Neural Networks made up of only operators efficiently supported across most architectures of ASICs. When a Arch-Net is produced, less common network constructs, like Layer Normalization and Embedding Layers, are eliminated in a progressive manner through label-free Blockwise Model Distillation, while performing sub-eight bit quantization at the same time to maximize performance. Empirical results on machine translation and image classification tasks confirm that we can transform latest developed Neural Architectures into fast running and as-accurate Arch-Net, ready for deployment on multiple mass-produced ASIC chips. The code will be available at https://github.com/megvii-research/Arch-Net.
In recent years, low-resource Machine Reading Comprehension (MRC) has made significant progress, with models getting remarkable performance on various language datasets. However, none of these models have been customized for the Urdu language. This work explores the semi-automated creation of the Urdu Question Answering Dataset (UQuAD1.0) by combining machine-translated SQuAD with human-generated samples derived from Wikipedia articles and Urdu RC worksheets from Cambridge O-level books. UQuAD1.0 is a large-scale Urdu dataset intended for extractive machine reading comprehension tasks consisting of 49k question Answers pairs in question, passage, and answer format. In UQuAD1.0, 45000 pairs of QA were generated by machine translation of the original SQuAD1.0 and approximately 4000 pairs via crowdsourcing. In this study, we used two types of MRC models: rule-based baseline and advanced Transformer-based models. However, we have discovered that the latter outperforms the others; thus, we have decided to concentrate solely on Transformer-based architectures. Using XLMRoBERTa and multi-lingual BERT, we acquire an F1 score of 0.66 and 0.63, respectively.
Estimating the quality of machine translation systems has been an ongoing challenge for researchers in this field. Many previous attempts at using round-trip translation as a measure of quality have failed, and there is much disagreement as to whether it can be a viable method of quality estimation. In this paper, we revisit round-trip translation, proposing a system which aims to solve the previous pitfalls found with the approach. Our method makes use of recent advances in language representation learning to more accurately gauge the similarity between the original and round-trip translated sentences. Experiments show that while our approach does not reach the performance of current state of the art methods, it may still be an effective approach for some language pairs.
This paper reports on the evaluation of Deep Learning (DL) transformer architecture models for Named-Entity Recognition (NER) on ten low-resourced South African (SA) languages. In addition, these DL transformer models were compared to other Neural Network and Machine Learning (ML) NER models. The findings show that transformer models substantially improve performance when applying discrete fine-tuning parameters per language. Furthermore, fine-tuned transformer models outperform other neural network and machine learning models on NER with the low-resourced SA languages. For example, the transformer models obtained the highest F-scores for six of the ten SA languages and the highest average F-score surpassing the Conditional Random Fields ML model. Practical implications include developing high-performance NER capability with less effort and resource costs, potentially improving downstream NLP tasks such as Machine Translation (MT). Therefore, the application of DL transformer architecture models for NLP NER sequence tagging tasks on low-resourced SA languages is viable. Additional research could evaluate the more recent transformer architecture models on other Natural Language Processing tasks and applications, such as Phrase chunking, MT, and Part-of-Speech tagging.
This paper proposes a tool for efficiently constructing high-quality parallel corpora with minimizing human labor and making this tool publicly available. Our proposed construction process is based on neural machine translation (NMT) to allow for it to not only coexist with human translation, but also improve its efficiency by combining data quality control with human translation in a data-centric approach.
Parallel corpus is a critical resource in machine learning-based translation. The task of collecting, extracting, and aligning texts in order to build an acceptable corpus for doing the translation is very tedious most especially for low-resource languages. In this paper, we present the efforts made to build a parallel corpus for Cebuano and Filipino from two different domains: biblical texts and the web. For the biblical resource, subword unit translation for verbs and copy-able approach for nouns were applied to correct inconsistencies in the translation. This correction mechanism was applied as a preprocessing technique. On the other hand, for Wikipedia being the main web resource, commonly occurring topic segments were extracted from both the source and the target languages. These observed topic segments are unique in 4 different categories. The identification of these topic segments may be used for the automatic extraction of sentences. A Recurrent Neural Network was used to implement the translation using OpenNMT sequence modeling tool in TensorFlow. The two different corpora were then evaluated by using them as two separate inputs in the neural network. Results have shown a difference in BLEU scores in both corpora.
The predominant approach for language modeling is to process sequences from left to right, but this eliminates a source of information: the order by which the sequence was generated. One strategy to recover this information is to decode both the content and ordering of tokens. Existing approaches supervise content and ordering by designing problem-specific loss functions and pre-training with an ordering pre-selected. Other recent works use iterative search to discover problem-specific orderings for training, but suffer from high time complexity and cannot be efficiently parallelized. We address these limitations with an unsupervised parallelizable learner that discovers high-quality generation orders purely from training data -- no domain knowledge required. The learner contains an encoder network and decoder language model that perform variational inference with autoregressive orders (represented as permutation matrices) as latent variables. The corresponding ELBO is not differentiable, so we develop a practical algorithm for end-to-end optimization using policy gradients. We implement the encoder as a Transformer with non-causal attention that outputs permutations in one forward pass. Permutations then serve as target generation orders for training an insertion-based Transformer language model. Empirical results in language modeling tasks demonstrate that our method is context-aware and discovers orderings that are competitive with or even better than fixed orders.
Machine translation (MT) system aims to translate source language into target language. Recent studies on MT systems mainly focus on neural machine translation (NMT). One factor that significantly affects the performance of NMT is the availability of high-quality parallel corpora. However, high-quality parallel corpora concerning Korean are relatively scarce compared to those associated with other high-resource languages, such as German or Italian. To address this problem, AI Hub recently released seven types of parallel corpora for Korean. In this study, we conduct an in-depth verification of the quality of corresponding parallel corpora through Linguistic Inquiry and Word Count (LIWC) and several relevant experiments. LIWC is a word-counting software program that can analyze corpora in multiple ways and extract linguistic features as a dictionary base. To the best of our knowledge, this study is the first to use LIWC to analyze parallel corpora in the field of NMT. Our findings suggest the direction of further research toward obtaining the improved quality parallel corpora through our correlation analysis in LIWC and NMT performance.
Recent research suggests that neural machine translation (MT) in the news domain has reached human-level performance, but for other professional domains, it is far below the level. In this paper, we conduct a fine-grained systematic human evaluation for four widely used Chinese-English NMT systems on scientific abstracts which are collected from published journals and books. Our human evaluation results show that all the systems return with more than 10\% error rates on average, which requires much post editing effort for real academic use. Furthermore, we categorize six main error types and and provide some real examples. Our findings emphasise the needs that research attention in the MT community should be shifted from short text generic translation to professional machine translation and build large scale bilingual corpus for these specific domains.
Speech-to-speech translation combines machine translation with speech synthesis, introducing evaluation challenges not present in either task alone. How to automatically evaluate speech-to-speech translation is an open question which has not previously been explored. Translating to speech rather than to text is often motivated by unwritten languages or languages without standardized orthographies. However, we show that the previously used automatic metric for this task is best equipped for standardized high-resource languages only. In this work, we first evaluate current metrics for speech-to-speech translation, and second assess how translation to dialectal variants rather than to standardized languages impacts various evaluation methods.
Simultaneous translation is a task in which translation begins before the speaker has finished speaking, so it is important to decide when to start the translation process. However, deciding whether to read more input words or start to translate is difficult for language pairs with different word orders such as English and Japanese. Motivated by the concept of pre-reordering, we propose a couple of simple decision rules using the label of the next constituent predicted by incremental constituent label prediction. In experiments on English-to-Japanese simultaneous translation, the proposed method outperformed baselines in the quality-latency trade-off.
We introduce a high-quality and large-scale Vietnamese-English parallel dataset of 3.02M sentence pairs, which is 2.9M pairs larger than the benchmark Vietnamese-English machine translation corpus IWSLT15. We conduct experiments comparing strong neural baselines and well-known automatic translation engines on our dataset and find that in both automatic and human evaluations: the best performance is obtained by fine-tuning the pre-trained sequence-to-sequence denoising auto-encoder mBART. To our best knowledge, this is the first large-scale Vietnamese-English machine translation study. We hope our publicly available dataset and study can serve as a starting point for future research and applications on Vietnamese-English machine translation.
This work explores the capacities of character-based Neural Machine Translation to translate noisy User-Generated Content (UGC) with a strong focus on exploring the limits of such approaches to handle productive UGC phenomena, which almost by definition, cannot be seen at training time. Within a strict zero-shot scenario, we first study the detrimental impact on translation performance of various user-generated content phenomena on a small annotated dataset we developed, and then show that such models are indeed incapable of handling unknown letters, which leads to catastrophic translation failure once such characters are encountered. We further confirm this behavior with a simple, yet insightful, copy task experiment and highlight the importance of reducing the vocabulary size hyper-parameter to increase the robustness of character-based models for machine translation.
Automatic Speech Recognition (ASR) systems have been gaining popularity in the recent years for their widespread usage in smart phones and speakers. Building ASR systems for task-specific scenarios is subject to the availability of utterances that adhere to the style of the task as well as the language in question. In our work, we target such a scenario wherein task-specific text data is available in a language that is different from the target language in which an ASR Language Model (LM) is expected. We use Neural Machine Translation (NMT) as an intermediate step to first obtain translations of the task-specific text data. We then train LMs on the 1-best and N-best translations and study ways to improve on such a baseline LM. We develop a procedure to derive word confusion networks from NMT beam search graphs and evaluate LMs trained on these confusion networks. With experiments on the WMT20 chat translation task dataset, we demonstrate that NMT confusion networks can help to reduce the perplexity of both n-gram and recurrent neural network LMs compared to those trained only on N-best translations.
We present a method for matching a text sentence from a given corpus to a given video clip and vice versa. Traditionally video and text matching is done by learning a shared embedding space and the encoding of one modality is independent of the other. In this work, we encode the dataset data in a way that takes into account the query's relevant information. The power of the method is demonstrated to arise from pooling the interaction data between words and frames. Since the encoding of the video clip depends on the sentence compared to it, the representation needs to be recomputed for each potential match. To this end, we propose an efficient shallow neural network. Its training employs a hierarchical triplet loss that is extendable to paragraph/video matching. The method is simple, provides explainability, and achieves state-of-the-art results for both sentence-clip and video-text by a sizable margin across five different datasets: ActivityNet, DiDeMo, YouCook2, MSR-VTT, and LSMDC. We also show that our conditioned representation can be transferred to video-guided machine translation, where we improved the current results on VATEX. Source code is available at https://github.com/AmeenAli/VideoMatch.
We consider the problem of multilingual unsupervised machine translation, translating to and from languages that only have monolingual data by using auxiliary parallel language pairs. For this problem the standard procedure so far to leverage the monolingual data is back-translation, which is computationally costly and hard to tune. In this paper we propose instead to use denoising adapters, adapter layers with a denoising objective, on top of pre-trained mBART-50. In addition to the modularity and flexibility of such an approach we show that the resulting translations are on-par with back-translating as measured by BLEU, and furthermore it allows adding unseen languages incrementally.
In order to achieve deep natural language understanding, syntactic constituent parsing is a vital step, highly demanded by many artificial intelligence systems to process both text and speech. One of the most recent proposals is the use of standard sequence-to-sequence models to perform constituent parsing as a machine translation task, instead of applying task-specific parsers. While they show a competitive performance, these text-to-parse transducers are still lagging behind classic techniques in terms of accuracy, coverage and speed. To close the gap, we here extend the framework of sequence-to-sequence models for constituent parsing, not only by providing a more powerful neural architecture for improving their performance, but also by enlarging their coverage to handle the most complex syntactic phenomena: discontinuous structures. To that end, we design several novel linearizations that can fully produce discontinuities and, for the first time, we test a sequence-to-sequence model on the main discontinuous benchmarks, obtaining competitive results on par with task-specific discontinuous constituent parsers and achieving state-of-the-art scores on the (discontinuous) English Penn Treebank.
Synthesis planning and reaction outcome prediction are two fundamental problems in computer-aided organic chemistry for which a variety of data-driven approaches have emerged. Natural language approaches that model each problem as a SMILES-to-SMILES translation lead to a simple end-to-end formulation, reduce the need for data preprocessing, and enable the use of well-optimized machine translation model architectures. However, SMILES representations are not an efficient representation for capturing information about molecular structures, as evidenced by the success of SMILES augmentation to boost empirical performance. Here, we describe a novel Graph2SMILES model that combines the power of Transformer models for text generation with the permutation invariance of molecular graph encoders that mitigates the need for input data augmentation. As an end-to-end architecture, Graph2SMILES can be used as a drop-in replacement for the Transformer in any task involving molecule(s)-to-molecule(s) transformations. In our encoder, an attention-augmented directed message passing neural network (D-MPNN) captures local chemical environments, and the global attention encoder allows for long-range and intermolecular interactions, enhanced by graph-aware positional embedding. Graph2SMILES improves the top-1 accuracy of the Transformer baselines by $1.7\%$ and $1.9\%$ for reaction outcome prediction on USPTO_480k and USPTO_STEREO datasets respectively, and by $9.8\%$ for one-step retrosynthesis on the USPTO_50k dataset.
Adapter layers are lightweight, learnable units inserted between transformer layers. Recent work explores using such layers for neural machine translation (NMT), to adapt pre-trained models to new domains or language pairs, training only a small set of parameters for each new setting (language pair or domain). In this work we study the compositionality of language and domain adapters in the context of Machine Translation. We aim to study, 1) parameter-efficient adaptation to multiple domains and languages simultaneously (full-resource scenario) and 2) cross-lingual transfer in domains where parallel data is unavailable for certain language pairs (partial-resource scenario). We find that in the partial resource scenario a naive combination of domain-specific and language-specific adapters often results in `catastrophic forgetting' of the missing languages. We study other ways to combine the adapters to alleviate this issue and maximize cross-lingual transfer. With our best adapter combinations, we obtain improvements of 3-4 BLEU on average for source languages that do not have in-domain data. For target languages without in-domain data, we achieve a similar improvement by combining adapters with back-translation. Supplementary material is available at https://tinyurl.com/r66stbxj
Recent work in simultaneous machine translation is often trained with conventional full sentence translation corpora, leading to either excessive latency or necessity to anticipate as-yet-unarrived words, when dealing with a language pair whose word orders significantly differ. This is unlike human simultaneous interpreters who produce largely monotonic translations at the expense of the grammaticality of a sentence being translated. In this paper, we thus propose an algorithm to reorder and refine the target side of a full sentence translation corpus, so that the words/phrases between the source and target sentences are aligned largely monotonically, using word alignment and non-autoregressive neural machine translation. We then train a widely used wait-k simultaneous translation model on this reordered-and-refined corpus. The proposed approach improves BLEU scores and resulting translations exhibit enhanced monotonicity with source sentences.