Kevin Duh

Bio: Kevin Duh is an academic researcher from Johns Hopkins University. The author has contributed to research in topics: Machine translation & Language model. The author has an hindex of 38, co-authored 205 publications receiving 5369 citations. Previous affiliations of Kevin Duh include University of Washington & Nara Institute of Science and Technology.

Lexicon Acquisition for Resource-Poor Languages Using Transductive Learning

Abstract: We investigate the problem of learning a part-of-speech (POS) lexicon for resource-poor languages. Developing a high-quality lexicon is often the rst step towards building a POS tagger, which is in turn the front-end to many NLP systems. We frame the lexicon acquisition problem as a transductive learning problem, and perform comparisons on three transductive algorithms: Transductive SVMs, Spectral Graph Transducers, and a novel Transductive Clustering method. We test on two datasets: dialectal Arabic (a resource-poor language) and Wall Street Journal with articially limited training data. For dialectal Arabic, we demonstrate that lexicon learning is an important task and leads to signicant improvements in tagging accuracy. For Wall Street Journal, we observe that transductive learning does not necessary lead to improvements in lexicon accuracy and present some preliminary analyses of results.

Approaching Sign Language Gloss Translation as a Low-Resource Machine Translation Task.

Abstract: A cascaded Sign Language Translation system first maps sign videos to gloss annotations and then translates glosses into a spoken languages. This work focuses on the second-stage gloss translation component, which is challenging due to the scarcity of publicly available parallel data. We approach gloss translation as a low-resource machine translation task and investigate two popular methods for improving translation quality: hyperparameter search and backtranslation. We discuss the potentials and pitfalls of these methods based on experiments on the RWTH-PHOENIX-Weather 2014T dataset.

An Empirical Analysis of Multiple-Turn Reasoning Strategies in Reading Comprehension Tasks

Abstract: Reading comprehension (RC) is a challenging task that requires synthesis of information across sentences and multiple turns of reasoning. Using a state-of-the-art RC model, we empirically investigate the performance of single-turn and multiple-turn reasoning on the SQuAD and MS MARCO datasets. The RC model is an end-to-end neural network with iterative attention, and uses reinforcement learning to dynamically control the number of turns. We find that multiple-turn reasoning outperforms single-turn reasoning for all question and answer types; further, we observe that enabling a flexible number of turns generally improves upon a fixed multiple-turn strategy. %across all question types, and is particularly beneficial to questions with lengthy, descriptive answers. We achieve results competitive to the state-of-the-art on these two datasets.

A Generalized Framework for Hierarchical Word Sequence Language Model

Abstract: Language modeling is a fundamental research problem that has wide application for many NLP tasks. For estimating probabilities of natural language sentences, most research on language modeling use n-gram based approaches to factor sentence probabilities. However, the assumption under n-gram models is not robust enough to cope with the data sparseness problem, which affects the final performance of language models. At the point, Hierarchical Word Sequence (abbreviated as HWS) language models can be viewed as an effective alternative to normal n-gram method. In this paper, we generalize HWS models into a framework, where different assumptions can be adopted to rearrange word sequences in a totally unsupervised fashion, which greatly increases the expandability of HWS models. For evaluation, we compare our rearranged word sequences to conventional n-gram word sequences. Both intrinsic and extrinsic experiments verify that our framework can achieve better performance, proving that our method can be considered as a better alternative for ngram language models.

An Analysis of Euclidean vs. Graph-Based Framing for Bilingual Lexicon Induction from Word Embedding Spaces.

Abstract: Much recent work in bilingual lexicon induction (BLI) views word embeddings as vectors in Euclidean space. As such, BLI is typically solved by finding a linear transformation that maps embeddings to a common space. Alternatively, word embeddings may be understood as nodes in a weighted graph. This framing allows us to examine a node's graph neighborhood without assuming a linear transform, and exploits new techniques from the graph matching optimization literature. These contrasting approaches have not been compared in BLI so far. In this work, we study the behavior of Euclidean versus graph-based approaches to BLI under differing data conditions and show that they complement each other when combined. We release our code at this https URL.

Automatic differentiation in PyTorch

Abstract: In this article, we describe an automatic differentiation module of PyTorch — a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototyping easier, PyTorch does not follow the symbolic approach used in many other deep learning frameworks, but focuses on differentiation of purely imperative programs, with a focus on extensibility and low overhead. Note that this preprint is a draft of certain sections from an upcoming paper covering all PyTorch features.

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Abstract: Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it provides an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several common benchmarks.

Pattern Recognition and Machine Learning

Abstract: Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Language Models are Few-Shot Learners

Abstract: Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Abstract: Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks.