K
Kevin Duh
Researcher at Johns Hopkins University
Publications - 205
Citations - 6391
Kevin Duh is an academic researcher from Johns Hopkins University. The author has contributed to research in topics: Machine translation & Parsing. 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.
Papers
More filters
Posted Content
A Call for Prudent Choice of Subword Merge Operations in Neural Machine Translation
TL;DR: This paper conducts a systematic exploration of different BPE merge operations to understand how it interacts with the model architecture, the strategy to build vocabularies and the language pair, and could provide guidance for selecting proper BPE configurations in the future.
Proceedings ArticleDOI
Comparing Pipelined and Integrated Approaches to Dialectal Arabic Neural Machine Translation
Pamela Shapiro,Kevin Duh +1 more
TL;DR: This work explores under which conditions it is beneficial to perform dialect identification for Arabic neural machine translation versus using a general system for all dialects.
Book ChapterDOI
Automated Development of DNN Based Spoken Language Systems Using Evolutionary Algorithms
TL;DR: This chapter explains the basic concepts and the neural network-based implementations of spoken language processing systems, and introduces the effort to automate the tuning of the system meta-parameters using evolutionary algorithms.
Proceedings Article
When Does Unsupervised Machine Translation Work
TL;DR: The authors conducted an extensive empirical evaluation using dissimilar language pairs, dissimilar domains, and diverse datasets and found that performance rapidly deteriorates when source and target corpora are from different domains and that stochasticity during embedding training can dramatically affect downstream results.
Posted Content
Query Expansion for Cross-Language Question Re-Ranking.
TL;DR: This work investigates expansions based on Word Embeddings, DBpedia concepts linking, and Hypernym, and shows that they outperform existing state-of-the-art methods on the cross-language question re-ranking shared task.